Signal, WhatsApp, and iMessage: Which Messaging App Is Most Secure?

I don’t know very much about what goes into war planning, but I assume that the communications infrastructure that supports that kind of thing is a solved problem for the government. There are secure telephone and video systems, as well as Secure Compartmented Information Facilities (SCIF) that allow the key players to review the most sensitive information about military activities. Typically, I assume, those sorts of conversations aren’t had using consumer messaging platforms on the Secretary of Defense’s iPhone. Also, I sort of assume that the people involved are smart enough and tech-savvy enough to notice that a journalist has entered the group chat. Apparently, not. There are a lot of questions raised by what is now certainly the most infamous group chat in the world, in which the Vice President, Secretaries of Defense and State, the Director of National Intelligence, CIA Director, and National Security Advisor were messaging about plans to bomb Houthi rebels in Yemen. We know about the chat because someone accidentally added Jeffrey Goldberg, the editor of The Atlantic. One question that a lot of readers might be wondering is just how secure the most popular messaging apps are. Here’s a rundown. Signal Signal, the app in question in this case, is end-to-end encrypted (E2EE). That means that messages are sent in an encrypted format and can only be read by the recipient. At the core of its encryption is the Signal Protocol, and open source protocol that allows for public inspection. That decreases the chances of hidden vulnerabilities. Signal also uses a form of encryption that ensures that even if a session key is compromised, previous messages stay encrypted. Signal also allows the most privacy since you don’t have to link a phone number to use the service (unlike other apps on this list). It also allows for contact verification so that you can ensure that the person you’re messaging is who they say they are. In general, Signal is widely considered the most secure consumer messaging app because third-parties can verify its security claims, and the company does not have access to metadata about your conversations. iMessage If you only send messages to other iPhone users, Apple’s iMessage platform is arguably the best and most secure option. Unlike Signal, Apple’s protocol is proprietary and not open for inspection by third-party security researchers. That makes it harder to verify that it is as secure as it claims, but Apple is well known for its commitment to security and privacy. One advantage is that Apple uses a 1:1 encryption model for group chats, which means that every message is encrypted individually for each member of the group. This is technically more secure than Signal’s Sender Key method, though it means that iMessage group chats are much more limited in terms of group size (due to the resources required for all of that individual encryption). Apple also says its encryption is designed for post-quantum computing. The idea is that eventually quantum computers will be able to break encryption easily enough to read protected messages, but Apple is designing its algorithm to resist those types of future capabilities. There are, however, two main drawbacks to Apple’s messaging platform. The first is that once you start messaging anyone with an Android device, it will fall back to RCS, or, worse, SMS—neither of which are encrypted within the Messages app. RCS supports E2EE, but Apple has not implemented the ability to send encrypted messages to Android devices. The other is that if you use iCloud backup for your messages, and aren’t using Advanced Data Protection, a copy of your messages is stored on Apple’s servers. While they are encrypted at rest, the company is able to turn them over if requested by law enforcement because it retains a key. WhatsApp WhatsApp uses the Signal Protocol (see above), meaning it offers a reliably secure form of protection for messages by default. One problem with WhatsApp is that, while the content of your messages may be encrypted, the metadata about the messages you send, and who you send them to, is not. That information is collected and stored by WhatsApp. Some people are also less than enthusiastic about using an app owned by Meta, which isn’t exactly known for its ability to keep its hands off of user data. It does, however, have the benefit of a massive user base, which means that there’s a good chance that the person you want to message with will be using WhatsApp. The app also has the best feature set for group messaging by far. Telegram To be clear, Telegram is not an E2EE messaging platform by default. Every regular message you send is encrypted in transit, and is encrypted as it is stored on Telegram’s servers, but that’s not the same thing as being encrypted so that only the recipient can read your message. This makes your messages vulnerable to anyone who has access to those servers. The app does allow you to create a “Secret Chat,” which is encrypted, and you can even set these to delete after a period of time. Still, if you care about protecting your text conversations, there are far better options on this list. Messenger Meta’s “other” messaging platform started rolling out E2EE last year, which should eventually put it on par with WhatsApp. The drawback here is that the rollout is happening over time, which means that not every user will immediately have it turned on by default. In addition, you might have some chats that are protected, and others that aren’t, and the average user isn’t going to know how to tell the difference. The bottom line It’s worth mentioning, however, that it does not matter how private or secure the encryption is on a messaging platform—if you include someone in a group chat and send a message to that group, they’re going to be able to read the message. Or, put another way, the problem here has nothing to do with encryption, and everything to do with human error. Most of these apps offer a secure form of E2EE for consumers, but there is no guarantee your messages will stay secret if you text them to a journalist. EXPERT OPINION BY JASON ATEN, TECH COLUMNIST @JASONATEN

Forecast: AI’s Rise Will Cut Search Engine Traffic, Affecting Advertising

A new report from the research firm Gartner, has some unsettling news for search engine giants like Google and Microsoft’s Bing. It predicts that as everyday net users become more comfortable with AI tech and incorporate it into their general net habits, chatbots and other agents will lead to a drop of 25 percent in “traditional search engine volume.” The search giants will then simply be “losing market share to AI chatbots and other virtual agents.” One reason to care about this news is to remember that the search engine giants are really marketing giants. Search engines are useful, but Google makes money by selling ads that leverage data from its search engine. These ads are designed to convert to profits for the companies whose wares are being promoted. Plus placing Google ads on a website is a revenue source that many other companies rely on–perhaps best known for being used by media firms. If AI upends search, then by definition this means it will similarly upend current marketing practices. And disrupted marketing norms mean that how you think about using online systems to market your company’s products will have to change too. AI already plays a role in marketing. Chatbots are touted as having copy generating skills that can boost small companies’ public relations efforts, but the tech is also having an effect inside the marketing process itself. An example of this is Shopify’s recent AI-powered Semantic Search system, which uses AI to sniff through the text and image data of a manufacturer’s products and then dream up better search-matching terms so that they don’t miss out on matching to customers searching for a particular phrase. But this is simply using AI to improve current search-based marketing systems. AI–smart enough to steal traffic More important is the notion that AI chatbots can “steal” search engine traffic. Think of how many of the queries that you usually direct at Google-from basic stuff like “what’s 200 Farenheit in Celcius?” to more complex matters like “what’s the most recent games console made by Sony?”–could be answered by a chatbot instead. Typing those queries into ChatGPT or a system like Microsoft’s Copilot could mean they aren’t directed through Google’s labyrinthine search engine systems. There’s also a hint that future web surfing won’t be as search-centric as it is now, thanks to the novel Arc app. Arc leverages search engine results as part of its answers to user queries, but the app promises to do the boring bits of web searching for you, neatly curating the answers above more traditional search engine results. AI “agents” are another emergent form of the tech that could impact search-AI systems that’re able to go off and perform a complex sequence of tasks for you, like searching for some data and analyzing it automatically. Google, of course, is savvy regarding these trends, and last year launched its own AI search push, with its Search Generative Experience. This is an effort to add in some of the clever summarizing abilities of generative AI systems to Google’s traditional search system, saving users time they’d otherwise have spent trawling through a handful of the top search results in order to learn the actual answer to the queries they typed in. But as AI use expands, and firms like Microsoft double– and triple-down on their efforts to incorporate AI into everyone’s digital lives, the question of the role of traditional search compared to AI chatbots and similar tech remains an open one. AI will soon impact how you think about marketing your company’s products and Search Engine Optimization to bolster traffic to your website may even stop being such an important factor. So if you’re building a long-term marketing strategy right now it might be worth examining how you can leverage AI products to market your wares alongside more traditional search systems. It’s always smart to skate to where the puck is going to be versus where it currently is. BY KIT EATON @KITEATON

OpenAI Says Using ChatGPT Can Make You Lonelier. Should You Limit AI Use at Work?

The dramatic increase in AI chatbot use created a new way for people and machines to interact, and that may be a problem. Researchers from market leading AI firm OpenAI and MIT’s Media Lab have concluded that using ChatGPT may actually worsen feelings of loneliness for people who use the chatbot all the time. If your company is using AI tools to speed up your worker’s days, this is definitely something to consider! While the results were presented with some nuance and subtlety, they fuel a narrative begun in 2023, when then Surgeon General Dr. Vivek Murthy warned that there was an “epidemic” of “loneliness and isolation” that was harming Americans’ health, and he partly blamed our digital era, noting to NPR that “we are living with technology that has profoundly changed how we interact with each other.” That doesn’t mean AI use should be banned at work, but it’s worth considering how long your employees are spending working with a chatbot. The authors of the joint study noted that their analysis found that while “most participants spent a relatively short amount of time chatting with the chatbot, a smaller number of participants engaged for significantly longer periods.” It’s these “power users” that may be experiencing the biggest impact. The authors noted that people who had “higher daily usage — across all modalities and conversation types — correlated with higher loneliness, dependence, and problematic use, and lower socialization.” Reporting on the study, Business Insider pointed out that in some ways this sort of investigation is always tricky because feelings of loneliness and social isolation often change from moment to moment, influenced by many factors. But to control for this the researchers measured both the survey participants’ feelings of loneliness and their actual level of socialization, to separate out when people really experience isolation from the feelings of loneliness. As with face-to-face interactions, tone was a big influence, the study concluded. When ChatGPT was instructed to react with flatter, more neutral interactions in a “formal, composed, and efficient” manner, the power users felt heightened loneliness. When the chatbot was told to be “delightful, spirited, and captivating” and reflect the user’s emotions, the users didn’t suffer this way. This makes sense: you wouldn’t necessarily feel listened to if, say, your AI conversations were as muted as a short discussion with a clerk at the department of motor vehicles, but you would feel differently if your AI coworker spoke more like the machine in Scarlett Johansson‘s movie Her. If your company has raced to embrace the promise of AI, does this mean you should rethink your AI tool usage, or, at the very least, worry about your staff mental health? The nuanced answer is probably not. Not yet, at least…but it’s definitely something to keep a weather eye on. AI technology is making its presence felt in the workplace, and it’s now capable of remarkably human-like level of interaction. Wired’s recent report about Google’s “scramble” to catch up with OpenAI will unsettle critics. After Google’s Gemini AI was perfected and released, one executive told Wired that “she had switched from calling her sister during her commutes to gabbing out loud with Gemini Live.” That’s a score for Google’s AI chops, but it’s easy to see how the change may impact that executive’s relationship with her family. For now, many AI tools exist more in the background in the workplace, and they’re less like chatting to a digital person than interacting with a smarter version of Microsoft’s old, much-loathed “clippy” digital assistant. For example, taking advantage of Microsoft’s AI Copilot suggestions to speed up writing an email in Outlook is unlikely to harm your workplace friendships in the way that gabbling to ChatGPT for 8 hours while sat at your desk might. At this point, connections are fraying in the increasingly digitized workplace. Workplace friendships are eroding, and the idea of a “workplace spouse” is fading. Tech company executives are pushing AI hard too, with Slack’s leadership imagining a near future when workers spend more time talking to AI agents at work than they do chatting with their colleagues. Last year, drug maker Moderna’s CEO said he was hoping his company’s bottom line would benefit from workers chatting to ChatGPT at least 20 times per day. Increased use of AI in the workplace could easily mean that many more people effectively become AI “power users,” triggering the kind of worry about loneliness that MIT’s researchers spoke of. No matter how much AI may drive up your employees’ efficiency, and boost your bottom line, it’s worth remembering that a ton of science shows that happy workers are better workers. BY KIT EATON @KITEATON

Microsoft’s AI Agents Aim to Make Cybersecurity Teams’ Work Easier

If you peek behind the curtain at a network defender’s workflow, you might see hundreds—if not thousands—of emails marked as potential spam or phishing. It can take hours to sift through the messages to detect the most urgent threats. When a data breach occurs, figuring out what vital information was stolen is a critical—but often challenging—step for investigators. Today, Microsoft announced a set of artificial intelligence agents aimed at making cybersecurity teams’ work a little easier. That could be good news for the many businesses large and small that use Microsoft 365 for their email, cloud storage, and other services. Agentic AI is a buzzy new term for AI systems that can take actions on behalf of a human user. One step up from generative AI chatbots, AI agents promise to do actual work, such as executing code or performing web searches. OpenAI recently launched Deep Research mode for ChatGPT, which can conduct multi-step web searches to research complex topics or make shopping recommendations for major purchases. Google has been rolling out its own AI agents built off the latest version of Gemini. A year ago, Microsoft launched Security Copilot, which introduced AI tools to its suite of security products: Purview, Defender, Sentinel, Intune, and Entra. Starting in April, users can opt in to having AI agents do specific tasks for them. Microsoft says the agents can help streamline the work of security and IT teams, which are facing both a labor shortage and an overwhelming volume of threats. Take phishing emails. In 2024, Microsoft says it detected 30 billion phishing emails targeting customers. At a company level, security teams often have to individually evaluate every potential phishing email and block malicious senders. A new phishing triage agent inside Defender scans messages flagged by employees to ensure that the most urgent threats are addressed first. Among the tasks the agent performs are reviewing messages for language that suggests a scam and checking for malicious links. The most dangerous emails go to the top of a user’s queue. Other messages might be deemed false positives, or simple spam. From there, the IT team can review a detailed description of the steps the agent took. The AI agent will suggest next steps—such as blocking all inbound emails from a domain associated with cybercriminals—and the human user can click a button to instruct the agent to perform those tasks. If an email was mistakenly marked as spam, there’s a field for the user to explain in natural language why that email should not have been flagged, helping train the AI to be more accurate going forward. Another AI agent helps prevent data loss—for example, looking for suspicious activity that might indicate an insider threat—and in the event of a data breach, helps investigators understand what information was stolen, whether a trade secret or customer credit card numbers. Other AI agents ensure new users and apps have the right security protocols in place, monitor for vulnerabilities, or analyze the evolving threat landscape a company faces. In each case, a user can look under the hood to see what steps the AI agent took in its investigation. The user can make corrections, or with the click of a button, tell the agent to complete the tasks it suggested, such as turning on multi-factor authentication for certain users, or running a software update. So far, the tools work across Microsoft services, such as Outlook, OneDrive and Teams, though integrations with third-party tools such as Slack or Zoom could be offered down the line. The tools also don’t take remediation steps without human approval. In the future, some of those tasks could also be automated. BY JENNIFER CONRAD @JENNIFERCONRAD

What Are AI Agents? Here’s How They Can Help You Get Stuff Done

OpenAI leader Sam Altman kicked off the year by declaring that 2025 would be when AI agents “join the workforce.” And he predicted that their addition will “materially change the output of companies.” It’s a sentiment shared by many AI industry luminaries, including Nvidia CEO Jensen Huang, who called agents “a multi-trillion-dollar opportunity.” (This type of technology is sometimes referred to as “agentic AI.”) But while the AI industry hypes up its shiny new toys, the rest of us are left with an important question: What the heck is an AI agent? What is an AI agent? OpenAI defines AI agents “as systems that independently accomplish tasks on behalf of users.” Olivier Godement, who heads up OpenAI’s API product team, says that an AI agent has been equipped with a specific purpose, a specific workflow to follow, and specific tools that enable it to take actions. That special purpose could be serving as a customer support specialist, and the actions it could take could be searching through an internal database, googling information, running a piece of code, or controlling a computer or internet browser. Essentially, agents are AI systems that can accomplish computer-based tasks for you. What’s the difference between an AI agent and a chatbot? Godement recommends thinking of agents as being the next evolution of chatbots. Apratim Purakayastha, general manager of Talent Development Solutions at online education platform Skillsoft, says that AI models become AI agents when they’ve been “enriched” with additional data and capabilities. You could direct an AI agent to continuously monitor prices for flights on a given date, or analyze job applications against an internal rubric. By combining the raw intelligence of AI models with data not included in its original training, says Purakayastha, a whole new world of opportunities arise. When they were originally released, chatbots like ChatGPT only existed as a channel for people to converse with AI models, but interactions were limited to the information the model had been trained on. Now, by activating the app’s Deep Research mode, users can instantly transform the chatbot into an AI agent, capable of searching the internet for minutes at a time and developing reports based on its findings. The AI model stays the same, but the added gadgets make it more useful. How have businesses used AI agents? According to Purakayastha, “gadgets” are actually a pretty good way to think about what makes AI agents different from AI models. Think about another notable agent: James Bond. Just like 007 receives a mission and gets equipped with gadgets specifically designed to help him carry out that mission, AI models are given a purpose and then equipped with the tools needed to carry out that purpose. A more mundane comparison is to an old-school travel agent, who books flights or hotel rooms for clients. One of the most successful examples of a company making use of an AI agent comes from software developer platform Replit. In 2024, the startup released an AI agent, built with an Anthropic Claude model, that can program applications from natural language prompts. Since the agent’s release, Replit has seen its revenue grow by over 10x. Other potential use cases for agents include providing customer support services, enabling more efficient data analysis, and, as recently demoed by Meta, empowering digital sales professionals. Can AI agents replace employees? One thing that AI agents aren’t? Virtual employees. Agents are intended to be used for very specific tasks, and shouldn’t be expected to act as a replacement for a human employee. Instead, says Purakayastha, agents should be thought of as a “virtual facet of an employee” or an “automated process.” A single agent should ideally only be responsible for completing a single step in a workflow, like approving or denying refunds by analyzing internal policy documents. You wouldn’t want to give the refund agent access to any unnecessary context or tools, because it would result in the agent becoming less focused on its specific task. Agents can be programmed to hand off control to each other, so you can essentially create a chain of agents that, when linked together, accomplish tasks. This process, which OpenAI’s Godement refers to as “separation of concerns,” also makes it easier to observe and optimize the effectiveness of agents, since they only need to be judged by their ability to complete a single step of a process. Can my business start using AI agents? AI agents are still in their infancy, and Godement cautions that businesses should “start small” when building agentic applications. Godement recommends identifying a workflow that you think could be more efficient, breaking down all the individual steps, and developing separate agents to handle each of those steps. “There will be a time when you can reinvent your whole product and company with agents,” he says, “but start small with one team of early adopters who are super passionate. You need one undeniable success story.” BY BEN SHERRY @BENLUCASSHERRY

OpenAI’s New Developer Tools Make Building AI Agents Easy

Earlier this week, OpenAI released two new solutions aimed at helping developers create AI-powered agents capable of going beyond just answering questions and actually taking actions, like approving refunds or buying plane tickets. Executives from OpenAI and cloud storage platform Box say these new development tools will make it much easier for businesses to take advantage of the cutting-edge tech. In a live streamed video on Tuesday, March 11, a group of OpenAI employees introduced the Agents software development kit (SDK) as well as the Responses application programming interface (API). The SDK gives developers a framework that enables applications built with OpenAI’s models to access additional tools and capabilities, like searching across files, parsing the internet, running code, or controlling a computer. The Responses API connects the company’s models to applications that require those new agentic capabilities. In essence, think of the SDK as a spell book and the API as a magic wand. The SDK gives developers the language needed to call upon these new powers, and the API channels those powers from their source (OpenAI) to an application. Developers give each agent a name, a set of instructions to follow (like “you are a customer support specialist … ”), and a defined set of tools that enables them to use specific capabilities or functions. In the live stream, the employees created an AI stylist agent as an example of how the new developer tools can be leveraged to create more useful AI applications. They gave the stylist agent two tools: the ability to search the internet, and access to a private database containing information about the OpenAI employees’ personal style. The agent used the insights from this data to recommend nearby stores with products that could match those preferences. To show how easily developers can upgrade their AI apps with new functionality, the OpenAI employees then created a second agent, assigned it to handle customer support, and gave it the ability to search through a database of previous orders and submit refund requests. In order to swap between the stylist and customer support agents, the team created a third agent, tasked solely with using contextual understanding to “hand off” control from one agent to the other. Olivier Godement, leader of OpenAI’s API business, says that within 24 hours of the SDK and API’s announcement, “several companies” had already pushed new products into production. OpenAI chief commercial officer Giancarlo Lionetti says he’s seen customers develop billing agents to send out invoices, financial analyst agents to compare market information found online with internal databases, and application agents for automating user enrollment of new programs and initiatives. It wasn’t impossible for developers to build agents before, says Godement, but it was a complex process that required multiple APIs. With OpenAI’s simplifying the process and creating a set of built-in tools, he says, it’s easier and faster than ever to equip AI models with agentic capabilities. One of the first new products built using the new API and SDK comes from Box, the cloud-based data storage platform. Box chief technology officer Ben Kus says that within 48 hours of receiving the development tools, his team created an agent that can connect proprietary data stored on the platform with OpenAI’s models, and use that information to accomplish various tasks. Imagine you have an e-commerce business and want to automate a customer service task, such as approving or denying a refund. You could create an agent, give it access to specific data on Box containing your company’s refund policy and customer order history, and then give it the ability to execute code approving or rejecting the claim based on the analyzed data. Box is so sure about agents’ ability to leverage data that the company is redesigning its internal search functions to be more agent-friendly. The data that a human may find useful when searching through their internal files isn’t necessarily as valuable to an AI agent, says Kus, so his team has been tailoring search integrations to ensure agents can obtain as much relevant information as possible. OpenAI’s Godement expects businesses to gradually become more ambitious with their agentic use cases over the next few months. Before long, he says, it could be common for an individual’s personal agent, with access to their identification and credit card info, to make transactions simply by conversing with a merchant’s agent. Basically, instead of scouring the internet yourself to buy a new sweater, your personal agent will handle the searching and shopping for you. A word of caution, though: Don’t overwhelm agents with an excess of information. With too much context, your agent may be a jack of all trades, but a master of none. Instead, Kus suggests thinking through how you’d accomplish a given task with humans, and then creating individual agents for every step of the workflow you’re hoping to automate. “Each agent has a job to be done,” says Lionetti, “and if you want it to do that job really well, it needs to be focused.” BY BEN SHERRY @BENLUCASSHERRY

AI is getting better at thinking like a person. Nvidia says its upgraded platform makes it even better

Nvidia on Tuesday revealed more details about its next artificial intelligence chip platform, Blackwell Ultra, which it says will help apps reason and act on a user’s behalf – two capabilities that would take AI beyond chatbots further into real life. Blackwell Ultra, which the company detailed at the its annual GTC conference, builds on Nvidia’s existing sought-after Blackwell chip. The additional computing power in the new Ultra version should make it easier for AI models to break complicated queries down into multiple steps and consider different options – in other words, reason, the company said. Demand for AI chips has surged in the wake of OpenAI’s ChatGPT in 2022, fueling a massive surge in Nvidia share prices. Its chips fuel the data centers that power popular, power-hungry AI and cloud-powered services from companies like Microsoft, Amazon and Google. But the arrival of Chinese tech startup DeepSeek – whose R1 model sent shockwaves through Wall Street for its reasoning capabilities and supposedly low cost – sparked speculation that expensive hardware may not be necessary to run high-performing AI models. Nvidia, however, appears to be skirting such concerns, as evidenced by its January quarter earnings in which it breezed past Wall Street’s expectations. Nvidia wants its chips to be central to the types of reasoning models that the Chinese tech startup helped popularize. It claims a DeepSeek R1 query that would have taken a minute and a half to answer on Nvidia’s previous-generation Hopper chip would only take 10 seconds with Blackwell Ultra. Cisco, Dell, HP, Lenovo and Supermicro are among the companies working on new servers based on Blackwell Ultra. The first products with Blackwell Ultra are expected to arrive in the second half of 2025. Being able to reason, or think through an answer before responding, will allow AI apps and agents to handle more complex and specific types of questions, experts say. Instead of just spitting out an answer, a chatbot with reasoning abilities could dissect a question and provide multiple, specific responses accounting for different scenarios. Nvidia cited an example of using a reasoning model to help create a seating arrangement for a wedding that takes into account preferences such as where to sit parents and in-laws and ensuring the bride is seated on the left. “The models are now starting to mimic a little bit of human-like behavior,” said Arun Chandrasekaran, an analyst covering artificial intelligence for market research firm Gartner. And it’s not just DeepSeek and OpenAI creating models that can reason. Google also updated its Gemini models with more reasoning capabilities late last year, and Anthropic introduced a hybrid reasoning model called Claude 3.7 Sonnet in February. Some experts also believe reasoning models could pave the way for “AI agents,” or AI assistants that can take actions rather than just answering questions. Companies like Google, Amazon and Qualcomm have been vocal about their visions for AI-powered helpers that can do things like book a vacation for you based on your preferences rather than just churning out answers to questions about flights and destinations. “What agentic AI excels at is multitasks,” said Gene Munster, managing partner at Deepwater Asset Management. “And being able to reason in each of those tasks is going to make the agents more capable.” By Lisa Eadicicco

If You Use AI Search at Work, Be Careful: A New Study Finds It Lies

With market leader Google leaning harder into AI-generated search results (after a stuttering start last year) and peer companies like OpenAI also trying out this innovative tech, it really does seem like AI is the future of online search. That will have implications for workers at pretty much any company, no matter the industry, because searching for information is such a fundamental part of the internet experience. But a new study from Columbia University’s Tow Center for Digital Journalism, featured in the Columbia Journalism Review, highlights that your staff needs to be really careful, at least for now, because AI search tools from several major makers have serious accuracy problems. The study concentrated on eight different AI search tools, including ChatGPT, Perplexity, Google’s Gemini, Microsoft’s Copilot, and the industry-upending Chinese tool DeepSeek; it centered around the accuracy of answers when each AI was quizzed about a news story, tech news site Ars Technica reported. The big takeaway from the study is that all the AIs demonstrated stunningly bad accuracy, answering 60 percent of the queries incorrectly. Not all the AIs were as bad as each other. Perplexity was incorrect about 37 percent of the time, while ChatGPT had a 67 percent error rate. Elon Musk’s Grok 3 model scored the worst, being incorrect 94 percent of the time—perhaps to no one’s surprise, given that Musk has touted the model as being limited by fewer safety constraints than rival AIs. (The billionaire also has a somewhat freewheeling attitude to facts and free speech.) Worse still, the researchers noted that premium, paid-for versions of these search tools sometimes fared worse than their free alternatives. It’s worth noting that AI search is slightly different to using an AI chatbot, which is more of a conversation. AI search typically sees the search engine trying to do the search for you after you type in your query, summarizing what it thinks are the important details from what it’s found online, so you don’t have to go and read the original article where the data comes from. But the problem here centers around the fact that, just like that one overconfident colleague who always seems to know the truth no matter what’s being discussed, these AI models just don’t like to admit they don’t know the answer to a query. The study’s authors noted that instead of saying “no” when they weren’t able to find reliable information in answer to a query on a news story, the AI frequently served up made-up, plausible-seeming, but actually incorrect answers. Another wrinkle detected by this study is that even when these AI search tools delivered citations to go alongside their search results (ostensibly so that users can then visit these source sites to double check any details, or to verify if the data is true) these citation links often led to syndicated versions of the content rather than the original publishers’ versions. Sometimes these links just led to web addresses that didn’t exist—Gemini and Grok 3 did this for more than half of their citations. Why should you care about this? The experiment was a bit niche, since it was based on news articles, and the researchers didn’t look deeply into the accuracy of AI search results for other content found online. Instead, they fed excerpts from real news pieces into the AI tools then asked them to summarize information, including the headline and other details. You should care for one simple reason. We know that AI can speed up some humdrum office tasks and boost employee efficiency. And it seems like AI search may become the norm, replacing traditional web searching, which can sometimes be a laborious task. But if your team is, for example, looking for background information to include in a piece of content you’re going to publish, or even looking for resources online before starting a new project, they need to be super careful about trusting the results that AI search tools deliver. Imagine if you published something on your company’s site only to learn that it was actually made up by an AI search tool that wasn’t prepared to say it didn’t know the actual answer. This is another version of the well-known AI hallucination problem, and it’s yet more proof that if you’re using AI tools to boost your company’s efforts, you definitely need savvy humans in the loop to check the AI’s output. BY KIT EATON @KITEATON

Sam Altman’s 8-Word Response to Meta’s Rumored ChatGPT Clone Is a Master Class

According to CNBC, Meta plans to launch a standalone app for AI during the second quarter of this year. There are few details available about what that might mean, except that it seems likely Meta will charge a monthly subscription, in line with existing players like ChatGPT, Anthropic’s Claude, and Google’s Gemini. If you’re the CEO of one of those companies, you’re probably thinking about how that might affect your business. After all, Meta is one of the largest companies in the world, with more than three billion users. But OpenAI CEO Sam Altman doesn’t seem all that worried. In a response posted on X, Altman suggested that his company will just “Uno reverse them,” and “do a social app” of its own. Look, I don’t think that OpenAI is launching some kind “ChatGPT With Friends” social app any time soon, but those eight words—”ok fine maybe we’ll do a social app”—are actually an interesting lesson. First of all, it should have been obvious to anyone paying attention that Meta would do this. Meta has never seen a good idea it wasn’t happy to steal from a competitor. In fact, almost every widely used feature in any of Meta’s apps started elsewhere. It’s kind of Mark Zuckerberg’s thing to see what the competition is doing and try to cut it off by just co-opting whatever feature people seem to love. You might argue it’s a gift, really. It has certainly been a successful model. Also, if Meta is serious about its AI chat ambitions, a standalone app seems like the only way to get there. I’m sure there are people who use Meta AI within the company’s various apps, but I’ve never talked to anyone who opens Instagram to chat with AI. They go to see what their friends are up to or get inspired by photos of other people’s lives, or to share funny videos. Still, this would represent an interesting change to Meta’s business model, which typically involves free apps and services that mostly serve as a way to target users with personalized ads. It makes sense that Meta would have to charge—AI inference is far more expensive than serving posts and showing ads—but this is a big shift. Meta is very much a consumer brand, which could expose generative AI tools to a much wider audience. I think a standalone Meta AI app could pose a real threat to ChatGPT in ways that other competitors have not. ChatGPT’s biggest advantage has been that it was first and its name is synonymous with AI for a lot of people. At the same time, it seems as though most of the people paying for ChatGPT or other similar AI tools are doing so for work or business reasons. I’m not sure how many of them will pay for another one. The most important thing here, however, is that Altman’s response is kind of perfect. No, not the part about OpenAI introducing a social app. I doubt that will actually happen—though I agree it would be funny, even if just on principle. Altman’s response does two things that serve as a master class to handling this kind of situation. First, and most important, it’s funny. Altman has mastered the art of reading the room and knowing his audience. On social media, Altman comes across as laid back and unaffected by the challenge to his business. I have no idea what he’s like in person, but this is the exact right way to respond in this environment. Second, Altman manages to highlight the biggest criticism you can make about Meta, which is the part where all of its best features are just copies of other people’s ideas. There’s not a lot you can do about that, except stay focused on what you’re doing instead of getting distracted by the threat of Meta coming for your business. If you look how quickly Meta was able to spin up Threads just by leveraging its existing social graph, there is no doubt in my mind that it could easily launch an app and have a few hundred million users in a short amount of time. If you’re Sam Altman, there is almost nothing you can do about that except continue making your product better. Oh, and have a sense of humor about the whole thing. EXPERT OPINION BY JASON ATEN, TECH COLUMNIST @JASONATEN

Google Says Its New AI Mode Will Help You Find Better, Deeper Information

Google has announced plans to expand generative AI’s presence in its search services with the launch of an experimental new chatbot, designed to help users “access more breadth and depth of information than a traditional search on Google,” according to a blog post announcing the new product. The company also announced updates to its controversial AI overviews, which provide a brief snapshot of key information related to a given Google search. The company has launched ‘AI Mode,’ an alternative to its traditional search interface. Unlike the default version of Google, AI Mode functions like a chatbot. Users can ask the chatbot a question, and the chatbot will use Google to search the web for information in order to quickly deliver an answer, complete with cited sources and links. In addition to conducting Google searches, AI Mode will integrate the company’s information systems, such as its stock market and weather trackers or shopping data. Google anticipates that one major use of AI Mode will be helping consumers find the right product for their specific needs and understand the differences between similar products. Google said in its blog post that it was expanding its AI search ambitions because “we’ve heard from power users that they want AI responses for even more of their search.” The company first attempted AI-augmented search with AI overviews, a feature in which an AI-generated answer to a Google search is posted at the top of the search results. Overviews debuted in May of 2024 to decidedly mixed reactions, but Google says it is one of the company’s “most popular services,” with over one billion global users. Notably, Google neglected to mention that overviews were automatically added to most users’ search experience, provided they were logged in to a Google account. AI Mode’s introduction signals that Google has a vision for the future of search that fully eliminates the need for users to choose from a set of websites, and instead acts more like a conversational assistant that can trawl the internet for you and only surface the best content. The move is a clear shot at OpenAI, which recently debuted a new feature that enables ChatGPT to search the internet, and Perplexity, an AI-powered search engine that is directly competing with Google. Google also announced that the overviews have been upgraded with Gemini 2.0, the latest version of the company’s series of AI models. Google says the new model will enable AI overviews to assist with a larger variety of questions and challenges, including “coding, advanced math, and multimodal queries.” The company added that teenagers can now use AI overviews, which going forward will appear even for users without Google accounts. Google analyzes search history data to estimate users’ ages. “As with any early-stage AI product,” Google said, “we won’t always get it right,” warning that AI Mode chatbot could appear biased on certain subjects. Currently, AI Mode is only available to paid Google One AI Premium subscribers. BY BEN SHERRY @BENLUCASSHERRY

Meet the Company Talking Big About Using AI to Render Private Wealth Managers Obsolete

Like many of AI’s biggest evangelists, Fahad Hassan believes the technology will transform entire industries, albeit with one distinction: Whereas other CEOs hope for a wave of so-called AI agents to reinvent areas like sales, Hassan is confident that bots will descend on the private wealth management space and irrevocably change the industry’s makeup. With that bullishness comes a grim prediction about the future employment of Certified Private Wealth Managers. “I tell them their job will not exist in the next five to ten years,” Hassan says with a smile. After co-founding the Mclean, Virginia-based Range in 2021, Hassan has helped build an investment management company that seeks to automate an industry that grew for decades on a bedrock of human relationships. He and his co-founder David Cusatis have strong votes of confidence from investors, including Google’s AI-focused fund Gradient Ventures, which contributed $12 million in Range’s Series A round in 2023, and also participated in its $28 million Series B last November. Hassan says more fundraising is coming: “We’re getting inbounds every 24 hours from AI-based VCs and other investors. We’re probably going to do a megaround this fall or this winter to accelerate our growth.” VCs are banking on their shared vision, one which is pretty rampant across Silicon Valley: The technological tide is turning and AI is making the upending of old norms inevitable. For Range’s purposes, this means that people don’t need to be on a first-name basis with a wealth advisor to make money. Trips to the golf course to build relationships and talk about ETFs are fast becoming a relic, thanks to AI agents, which work around the clock. “Stock brokers are a really good example where people had traded millions of dollars on the phone and knew a person. And today, you don’t do that. You don’t even think about it,” Hassan argues. Stock brokers’ demise has been noted before, although they still exist, often just under different titles like wealth managers, or financial advisers. The biggest question hanging over Range’s gambit is whether a platform that seeks to grow people’s money through new technology can earn enough public trust to unseat major financial institutions. How will it be regulated? Perhaps lightly, if Big Tech’s incursion into the federal government provides a clue: Many of President Donald Trump’s prominent backers hail from the tech industry, including the venture capitalists Ben Horowitz and Marc Andressen, whose firm, Andreessen Horowitz, has been praising AI-driven investment strategies since 2023. Hassan says it’s “the Wild West.” For now, Range isn’t even offering AI tools for its clients, making Hassan’s promises of sweeping automation sound perhaps a little blustery. “We’re still in the very, very early days of what a good AI user experience looks like,” says Cusatis, a software engineer who holds the title of chief architect at Range. For now, Range provides three pricing tiers that start at $2,655 annually. Unlike other private investment managers, Range doesn’t bill based on a percentage of a client’s assets. So far, the company’s custom-built AI tools, which both founders say are built on existing major Large Language Models, are used by Range’s team of wealth advisors to help tailor strategies for clients’ needs. Range is targeting a mid-year release of its first client-facing AI programs, and plans to keep its human team of advisors intact. The company’s ambition to automate jobs out of existence doesn’t apply to its own teams: “Range will have wealth managers, and it will always be a service they provide, if people want it,” a spokesperson tells Inc. The cautionary period makes sense. Some of the splashier rollouts in commercial AI have had major hiccups. Google’s AI overhaul of search famously recommended glue on pizza, and had a rocky start in general. Self-driving vehicles have also crashed, prompting scrutiny from regulators. So the question remains: Why should people trust a similar technology with their most important financial decisions? Alejandro Lopez-Lira, a professor of finance at the University of Florida, has been experimenting with how Large Language Models can be put to this test. There are some advantages, he says, in that an application like ChatGPT can absorb and synthesize far more information than any human can. But when it comes to financial planning over a long period, can an AI account for all the unforeseen variables that might arise and affect your portfolio? “It’s extremely hard to test the performance of any of these AI [investment] strategies, because they have basically memorized the whole internet up until very, very recently,” Lopez-Lira says. ChatGPT, for instance, can synthesize immense data, but only to a point: “For the latest ChatGPT models, the cutoff date is 2024. So if you want to use that model, you only have less than one year to simulate the performance of the strategy,” he explains. There is also the question of regulating autonomous investments. Algorithmic trading, or buying and selling securities on an automated basis, is something hedge funds have done for years, regulated by the Securities and Exchange commission and the Financial Industry Regulatory Authority, a non-profit that issues certifications. But agentic AI in private wealth management is a new area that isn’t regulated yet. “It’s a little bit of a gray area, for sure,” Lopez-Lira says. Hassan claims that’s to Range’s advantage. “Our vision is to keep doing what we’re doing growing as fast as we’re growing, and candidly, lobby for the set of laws that will come around AIs giving you advice and what that looks like.” he says. BY SAM BLUM, @SAMMBLUM

Anthropic’s Newest AI Wants to Be a Pokémon Master. Here’s Why That’s a Big Deal

On Monday, Anthropic released Claude 3.7 Sonnet, the company’s most capable AI model yet. It also revealed a new capability with implications for the business world: Claude can now play Pokémon, and it’s pretty good at it, at least for an AI. In a blog post detailing the new Claude model, Anthropic wrote that a small internal team had created an interface that enabled Claude to play Pokémon Red, the original Pokémon game released on the Nintendo Game Boy way back in 1996. So, why teach an AI model to play Pokémon? David Hershey, a member of Anthropic’s technical team, tells Inc. that staffers were inspired by a YouTube video in which an original reinforcement learning model was trained to play Pokémon, so they created a virtual environment in which Claude could attempt to play the game. Eventually, around June 2024, Hershey (a self-proclaimed Pokémon fan) took up the idea as a side project, first using it to test the capabilities of Claude 3.5 Sonnet, the new model at the time. He found that while earlier versions of Claude would immediately get stuck, Claude 3.5 could progress further, successfully catching a Pokémon and leaving the starting area of Pallet Town. For the uninitiated, the goal of Pokémon Red is to catch adorable creatures, train them by battling against non-playable characters, and win badges from powerful enemies called Gym Leaders. Pokemon is, of course, wildly popular, and is considered the highest-grossing media franchise of all time. To keep his coworkers, many of whom are also Pokémon fans, up-to-date on his efforts, Hershey started a dedicated Slack channel in which Anthropic employees could monitor Claude’s Pokémon journey. Every time Claude successfully caught a Pokémon or won a battle, more and more people would join the Slack channel, says Anthropic product research lead Diane Penn. The project developed a cult following within the company. Video games have long been used as a method for gauging an AI model’s ability. In the early days of OpenAI, staffers spent years training models to play online multiplayer game DOTA 2. “Defining success is hard,” says Hershey, “But video games happen to be structured in a way where progress is often measurable and linear.” Other games have been important over the decades as AI, or just thinking machines in general, developed. In 2016, Google DeepMind’s AlphaGo beat one of the world’s highest-ranked players of the ancient game Go, and in 2011, IBM’s Watson system beat Jeopardy champions Ken Jennings and Brad Rutter at their own game. And in 1997, IBM’s Deep Blue beat Garry Kasparov at chess. When Anthropic released an updated version of Claude 3.5 Sonnet in September, the model saw slight improvement, but the real breakthrough came in the form of Claude 3.7 Sonnet, the new model released this week. While the previous model got stuck in Viridian Forest, an early area in the game, 3.7 Sonnet was able to go further, collecting three badges from Pokémon gym leaders. So how is Claude able to improve itself? Hershey says that Claude’s new Pokémon skills are the direct result of a new feature called “extended thinking,” which enables the model to take additional time to “think” through how to solve a problem, instead of immediately generating a response. Hershey says a common complaint he’s heard from Anthopic’s customers is that earlier versions of Claude would make a false assumption and then struggle to reverse course. But because of its improved thought process, the new Claude is able to more effectively pivot and try new strategies, meaning it doesn’t get stuck nearly as often as earlier versions. According to Penn, Anthropic decided to include Hershey’s Pokémon benchmark in Claude 3.7 Sonnet’s announcement because the company is slowly moving away from traditional benchmarks in favor of more “accessible” tests that can be understood by a larger group of people. “We’re at a point where evaluations don’t tell the full story of how much more capable each version of these models are,” Penn says. Penn says the benchmark demonstrates Claude’s ability to intelligently make a plan and adapt with new strategies when it runs into a problem. For companies looking to use AI on complex tasks like conducting high quality research or complex financial analysis, the benchmark is proof that models can improve their performance by using reasoning capabilities. By using Pokémon progression as a benchmark, Anthropic is able to educate an entirely new audience about Claude’s capabilities. After learning that the benchmark would be included in the announcement, Hershey and a small team hustled to quickly create an ongoing livestream on Twitch, in which anyone can watch Claude attempt to catch ‘em all. Some users have even said in the livestream’s chat that they were inspired to subscribe to Anthropic’s $18 per month Claude Pro service. Even with its enhanced capabilities, the model is still far from becoming a Pokémon Master, say Penn and Hershey. As of Thursday afternoon, the model had been stuck in Mt. Moon, an early game area that’s notoriously tricky for kids, for over 27 hours. Viewers of “Claude Plays Pokémon” were especially delighted when Claude named its rival character Waclaud, a reference to Super Mario Bros villains Wario and Waluigi, but that actually wasn’t a decision made by the model. “It was in the system prompt,” admits Hershey. Before the livestream was launched, he says, “we ran an internal poll on what we should name the rival, so Waclaud is just a small easter egg from our internal culture at Anthropic.” Could the project’s popularity result in Pokémon becoming a new standardized benchmark for AI? Hershey isn’t quite sure, but based on the response his project has received online, he wouldn’t be surprised to see other AI labs use video games as benchmarks more often. “It’s just a great way to see progress over a long period of time,” he says, “and we’re definitely not the only people who think that’s important.” BY BEN SHERRY @BENLUCASSHERRY

Can AI Startups Dethrone Google Chrome in the Web Browser Wars?

A new report from the research firm Gartner, has some unsettling news for search engine giants like Google and Microsoft’s Bing. It predicts that as everyday net users become more comfortable with AI tech and incorporate it into their general net habits, chatbots and other agents will lead to a drop of 25 percent in “traditional search engine volume.” The search giants will then simply be “losing market share to AI chatbots and other virtual agents.” One reason to care about this news is to remember that the search engine giants are really marketing giants. Search engines are useful, but Google makes money by selling ads that leverage data from its search engine. These ads are designed to convert to profits for the companies whose wares are being promoted. Plus placing Google ads on a website is a revenue source that many other companies rely on–perhaps best known for being used by media firms. If AI upends search, then by definition this means it will similarly upend current marketing practices. And disrupted marketing norms mean that how you think about using online systems to market your company’s products will have to change too. AI already plays a role in marketing. Chatbots are touted as having copy generating skills that can boost small companies’ public relations efforts, but the tech is also having an effect inside the marketing process itself. An example of this is Shopify’s recent AI-powered Semantic Search system, which uses AI to sniff through the text and image data of a manufacturer’s products and then dream up better search-matching terms so that they don’t miss out on matching to customers searching for a particular phrase. But this is simply using AI to improve current search-based marketing systems. AI–smart enough to steal traffic More important is the notion that AI chatbots can “steal” search engine traffic. Think of how many of the queries that you usually direct at Google-from basic stuff like “what’s 200 Farenheit in Celcius?” to more complex matters like “what’s the most recent games console made by Sony?”–could be answered by a chatbot instead. Typing those queries into ChatGPT or a system like Microsoft’s Copilot could mean they aren’t directed through Google’s labyrinthine search engine systems. There’s also a hint that future web surfing won’t be as search-centric as it is now, thanks to the novel Arc app. Arc leverages search engine results as part of its answers to user queries, but the app promises to do the boring bits of web searching for you, neatly curating the answers above more traditional search engine results. AI “agents” are another emergent form of the tech that could impact search-AI systems that’re able to go off and perform a complex sequence of tasks for you, like searching for some data and analyzing it automatically. Google, of course, is savvy regarding these trends, and last year launched its own AI search push, with its Search Generative Experience. This is an effort to add in some of the clever summarizing abilities of generative AI systems to Google’s traditional search system, saving users time they’d otherwise have spent trawling through a handful of the top search results in order to learn the actual answer to the queries they typed in. But as AI use expands, and firms like Microsoft double– and triple-down on their efforts to incorporate AI into everyone’s digital lives, the question of the role of traditional search compared to AI chatbots and similar tech remains an open one. AI will soon impact how you think about marketing your company’s products and Search Engine Optimization to bolster traffic to your website may even stop being such an important factor. So if you’re building a long-term marketing strategy right now it might be worth examining how you can leverage AI products to market your wares alongside more traditional search systems. It’s always smart to skate to where the puck is going to be versus where it currently is. BY KIT EATON @KITEATON

Slack Imagines a Future Workplace Where You Chat More With AIs Than With Your Colleagues

The folks behind the messaging app Slack know a thing or two about how workers communicate with one another and their bosses. At least 750,0000 organizations around the world rely on it for their workplace communications. So when the company’s chief marketing officer makes a prediction about the future of workplace comms, it’s worth paying attention — and, boy, does Ryan Gavin have a doozy of an idea. In conversation with news outlet Axios, Gavin predicted that the rise of AI agents will transform workplaces, and that staff may soon talk to AIs more than to their human co-workers. AI agents have been hailed by many experts as the first truly useful tools that AI may provide, and possibly the next big thing in this technology revolution. Even OpenAI’s CEO Sam Altman is on board with this notion, and his company’s upcoming agent Operator may even prove to be the first AI gizmo that transforms the average workplace. Agents are more powerful than the ask-then-answer AI model chatbots use because they can actually perform actions in a digital environment, like filling in forms on a website automatically, or even taking control of your computer’s mouse and using apps on the desktop. Axios reminds us that Salesforce, which owns Slack, has been promoting its own AI agent systems, which are apparently already capable of acting like sales reps. But instead of being innovations looming in a far-off future, Gavin said he anticipates these agent systems achieving everyday use in many workplaces sooner rather than later. “I think that right now people are underestimating just how much the world of work is about to change,” he told Axios, putting a timeline on the transformation brought by AI agents as just “three or four or five years.” By then he said he imagined he could be talking to agents “as much, if not more than I’m talking to my human colleagues today.” This projection may unsettle AI critics who worry the tech will seriously disturb the way that humans interact with each other in the office, possibly contributing to worker burnout or the erosion of human relations, and even displace people from their jobs. But Gavin’s prediction aligns with numerous other expert views that suggest that AI’s will augment workers’ office skills, rather than replace them outright. Picture the scene if “every single employee had a human resources agent that sat right alongside them in Slack” Gavin said. As Axios noted, AI co-workers like this have the added benefit that they are easier to train than people, they may be cheaper to “employ,” they don’t ask for raises, and they won’t strike or quit. Gavin’s words brush over the obvious issue that workers often have an existential dislike of office human resources departments. Couple that with the notion that a computer-based company representative is digitally watching over your shoulder as you work, and the idea may worry people who already think that workplace surveillance solutions and worker time and task tracking are already far too Orwellian. That said, it’s easy to imagine an AI agent co-worker that would be very useful — it could serve up people’s contact info automatically when you’re planning a task, andf it could even fill in timesheets for you or look up specific company information, like financial data, when you’re putting together a presentation. How this will actually play out in the typical workplace of tomorrow is anyone’s guess, of course. Gavin’s point of view is merely one of many diverse perspectives, and seems centered more around digital messaging chats than actually talking to co-workers in person — no one is suggesting that office water cooler gossip will go away. But Gavin’s prediction of ubiquitous AI co-workers even aligns with recent data showing that employees now shun deep and lasting friendships with co-workers, since it suggests a digital colleague may take up some of this void. Supporters of AI use will also point out that some research suggests letting staff use AI in the workplace can actually boost their happiness. BY KIT EATON @KITEATON

Robot Startup Figure Reveals an AI Breakthrough Called Helix

Figure, a Silicon Valley-based startup that builds humanoid robots, has announced Helix, a proprietary AI model that the company says enables robots to more easily pick up objects, collaborate on tasks with other robots, and smoothly control their entire upper bodies. The launch comes weeks after Figure publicly ended a deal to collaborate on AI models with OpenAI, and after the company raised $1.5 billion. In a blog post, Figure wrote that Helix’s launch marks a major step in bringing humanoid robots into “everyday home environments.” Unlike previous versions of its models, Figure says that Helix doesn’t need to be fine-tuned or trained on hundreds of demonstrations in order to perform specific tasks. Instead, a new design allows the robots to interact with objects that they’ve never encountered before. The company says that Figure robots equipped with Helix “can pick up virtually any small household object.” In a video example of Helix’s capabilities, a man asks two Figure-made robots, both equipped with the Helix AI model, to put away a bag of groceries. After looking around, the robots slowly opened up a nearby fridge, put most of the food away, and placed an apple in a serving bowl. Up until recently, Figure had been in a partnership with OpenAI in which the two firms collaborated to customize OpenAI’s models so they could be used by Figure’s robots. But on February 4, Figure founder Brett Adcock announced on X that he had decided to pull his company out of the agreement. He said that Figure had “made a major breakthrough on fully end-to-end robot AI, built entirely in-house,” a clear nod to Helix’s imminent reveal. Adcock said that the company has been working on Helix for over a year. It’s been a busy month for Figure. On February 14, Bloomberg reported that the company was discussing a new funding round with firms Align Venture and Parkway Venture Capital that would raise $1.5 billion at a $39.5 billion valuation. Bloomberg said that much of investors’ enthusiasm comes from a recent breakthrough that has sped up Figure’s timeline for penetrating the home market with its robots, implying that investors got an early look at Helix. Figure isn’t the only company trying to make humanoid robots happen. Just last week, Reuters reported that Meta is establishing a new division that will create AI-powered humanoid robots, and humanoid robotics startup Apptronik raised $350 million, with one of its investors and partners being Google DeepMind. Also in the race is Tesla, with its “Optimus” robots, and rumors are swirling that Apple could enter the fray as well. BY BEN SHERRY @BENLUCASSHERRY

How Google’s New AI Co-Scientist Tool Gives Us a Taste of Tomorrow’s Workplace

Google, like many other big tech names, has already released numerous AI tools — some more generic, designed to help with a wide range of tasks, some of which are tailored to the specific needs of specialized users. Its latest effort is definitely in the latter category: the new “AI co-scientist” system, built on its Gemini 2.0 AI model, is specifically designed to “aid scientists in creating novel hypotheses and research plans.” That sounds like a very niche market—and it is. But it’s also likely to be the tip of the AI-as-coworker iceberg. In a post announcing the tool, Google explained how the new system would be used in a research setting. Essentially a scientist who has a specific topic to investigate — like, say, discovering a new drug to tackle a particular disease — would input that into the tool using natural language. The AI would then reply, much like any other chatbot, with a useful output — in this case a hypothesis that the scientist can then test to either validate or invalidate their theory. The tool also does some of the work that goes into starting a new experiment by summarizing published literature about the topic, and suggesting an experimental approach. Google’s blog post explains that the tool is actually a “multi-agent” system, tapping into what many think may be the next big thing in AI innovations. Using Gemini’s ability to reason, synthesize data and perform long-term planning, the tool roughly models the actual intellectual process scientists use when tackling a novel problem—the scientific method. In this case Google’s system uses four AI agents called Generation, Reflection, Ranking and Evolution, refining its answers over and over in what Google calls a “self-improving cycle of increasingly high-quality and novel outputs.” Essentially the tool cycles through lots of different ideas, checking how good they are and then spitting out what it thinks is the best output. Google is very careful to note that the tool is designed to be a scientific collaborator, to “help experts gather research and refine their work,” and it’s not meant to “automate the scientific process.” What this means is that the AI co-scientist isn’t designed to replace scientists, but instead may inspire researchers with novel ideas or otherwise speed up the process of investigating a thorny physics problem, or tackling a biological issue like antimicrobial resistance. The pros and cons of AI-assisted science In a previous career I worked with plenty of high-tech scientific machinery, from particle accelerators to complex computer-controlled lab equipment, using everything at my disposal to help advance my research. From this experience I can say Google’s AI tool would’ve been invaluable, saving me hours of time looking up material online and in physical texts, as well as when it came to thinking up clever ways to “break in” to a particular physics problem — the hypothesis formation and testing process at the core of scientific progress. It also seems likely other scientists will race to adopt a tool like this, because it would free up valuable time to do actual real-world experiments. For now the AI science tool is only available through a Google-led pilot testing program, to help “evaluate its strengths and limitations in science and biomedicine more broadly,” before a wider launch. I can foresee a couple of issues that may hold some researchers back, however. More traditional scientists may be reluctant to trust such a revolutionary new model for carrying out research — even if it comes at the expense of seeing their rivals embrace the technology. Scientists whose research is politically sensitive, or perhaps secret, may not trust an AI, simply because of the known issues of AI data “leakage,” where information put into the AI as queries can then emerge later on when a different user types in a prompt. Some researchers may thus be forbidden from using this sort of AI at all. Scientific research is also a highly creative process that requires feats of imagination and insight to create original work. Some researchers may be reluctant to hand over this part of the process to a machine. Nevertheless, if Google’s tool really works (and the blog post includes several examples where it’s been tested out in the real world, including helping researchers looking at liver fibrosis) scientists of all stripes may embrace having an AI coworker in the lab. In this way, the tool gives us a hint of how AI may penetrate into many different types of workplace over time, as more and more specialized AI systems are developed by Google and rival firms. The idea even taps into the thorny “will AI steal my job?” issue — with this particular example suggesting that no, it won’t: instead AI will kind of ride along with you as you work through your day, helping you as you need it to. BY KIT EATON @KITEATON

This Chinese AI Bet Has Outperformed Magnificent Seven Names Like Meta and Google This Year

Magnificent Seven who? That’s what some U.S. investors might be asking themselves after Alibaba’s stellar earnings pushed its U.S.-listed stock more than 8 percent higher Thursday, bringing its year-to-date gains above 60 percent. The strong showing reinforced its position among investors’ favorite AI bets, particularly in comparison to its U.S. peers like Meta, Alphabet and Microsoft. It’s worth highlighting, too, that intrigue around Alibaba has spiked this week, with its co-founder Jack Ma making a rare public appearance Monday with Chinese President Xi Jinping. Here’s how the company performed in the quarter to December 31: Earnings per share: $2.93, above expectations for $2.66 Revenue: 280.15 billion yuan, above LSEG expectations for 279.34 billion yuan Net income: 48.945 billion yuan, above LSEG expectations for 40.6 billion Notably, Alibaba’s Cloud Intelligence Group saw its sales growth surge 13% compared to a year ago to 31.742 billion yuan. “This quarter’s results demonstrated substantial progress in our ‘user first, AI-driven’ strategies and the re-accelerated growth of our core businesses,” Alibaba chief executive Eddie Wu said in a statement. Alibaba has proven largely unaffected by the recent DeepSeek scare that throttled US tech stocks. Only a week ago, Alibaba announced a partnership with Apple to integrate Alibaba’s AI into iPhones in China. Indeed, Bloomberg Intelligence analysts wrote in a note Thursday that the company’s decisions through 2024 have panned out as effective strategies to fend off domestic competition too from Huawei, Tencent, Baidu and others. It’s also better positioned than its rivals, in Bloomberg Intelligence’s view, to withstand headwinds from a potential US-China trade war. That said, Alibaba’s margins could see weaker growth should the company continue to push for revenue growth. “Alibaba is likely to sacrifice some margin growth in e-commerce, logistics and cloud through the fiscal year ending March 2026 as it focuses on driving up revenue,” said Bloomberg Intelligence senior industry analyst Catherine Lim. “The company aims to enlarge the market shares of these businesses, in mainland China and overseas. That could fuel up-front expenditure on technology-led process upgrades, AI-related development and logistics solutions in the next 13 months.” As a side note, the Wall Street Journal reported Thursday that GameStop’s billionaire CEO Ryan Cohen raised his personal stake in Alibaba to roughly 7 million shares worth about $1 billion. Cohen, who is steering a company with a rising stock in his own right, seems to like the AI play. “The AI era presents a clear and massive demand for infrastructure,” Alibaba’s Wu said on the earnings call. “We will aggressively invest in AI infrastructure. Our planned investment in cloud and AI infrastructure over the next three years is set to exceed what we have spent over the past decade.” BY PHIL ROSEN, CO-FOUNDER AND EDITOR, OPENING BELL DAILY @PHILROSENN

The AI Energy Score attempts to quantify the environmental harm of AI models. Here’s how to use it.

Artificial Intelligence is notoriously energy-hungry. Now, a new tool from Salesforce is attempting to quantify the toll AI models are taking on the planet. “AI models vary dramatically in terms of their environmental impact, from big general models to small domain-specific models,” says Boris Gamazaychikov, head of AI sustainability at Salesforce. “What the AI Energy Score is aiming for is to develop a standardized way to actually start measuring these things.” AI Energy Score, launched Monday by Salesforce in collaboration with Hugging Face, Cohere, and Carnegie Mellon University, assigns a star rating to assess the energy efficiency of various AI models relative to other models. The ratings are available for 10 distinct tasks, from text and image generation to summarization and automatic speech recognition. A five-star rating is the most energy-efficient, whereas a one-star rating is least. It also shares a number estimate of the quantity of GPU energy consumed per 1,000 search queries in Watt hours. As of Feb. 10, DistilGPT2, a model developed by Hugging Face that is meant to be a “faster, lighter version of GPT-2,” was the most energy efficient, consuming 1.31 Wh of energy per 1,000 text generation queries. Alongside an interactive landing page with more information, the site also has a label generator that Gamazaychikov hopes developers will use to showcase their scores and drive awareness. For now, the tool only rates open-source models hosted on Hugging Face, a platform for data scientists, AI developers, and enthusiasts. That includes some models from Meta, Microsoft, Mistral, and StabilityAI. Salesforce also submitted its own open- and closed-source models for rating. Closed-source, or proprietary models, which include many big names such as OpenAI’s GPT-3 and 4, Anthropic’s Claude, and Google’s Gemini, are not readily available. But Gamazaychikov says there is a way for developers of these proprietary models to conduct the analysis securely and then submit it if they so choose. “By making a really easy to use, clear, standardized approach, we’re reducing any kind of friction between those leading companies being able to not do this,” Gamazaychikov says, adding that the tool was developed using feedback from academia, government officials, and AI companies. “We’re also hoping to increase some pressure, showing that it’s important to disclose this type of information.” Although Salesforce welcomes individuals to experiment with the tool to get a sense of how their use of AI models could take a toll on the environment, Gamazaychikov says “the primary audience is probably enterprises, or those that are integrating these types of models within their products.” Furthermore, a tool like AI Energy Score could help companies calculate indirect greenhouse gas emissions throughout the value chain. The launch of AI Energy Score follows Salesforce’s debut last fall of Agentforce, which allows customers to build or customize AI-powered agents that automate tasks or augment employee labor. Gamazaychikov says Salesforce also hopes to inspire greater change by giving regulators a jumping-off point for assessing the emissions-load of AI. He also envisions promoting the use of smaller, local models at a time when large reasoning models, which could consume even more energy than current large language models, are in development. A 2024 study from Goldman Sachs found that AI could drive up the energy demand of data centers by 160 percent by the end of the decade, even as the growing global population exerts additional pressures on the world’s energy needs. This comes at a time when leaders are contending with the worsening effects of climate change, prompting big tech companies to look to renewable energy and innovations in nuclear power to satisfy the growing demand. Want to try it out? Head on over to huggingface.co/AIEnergyScore and give it a whirl. BY CHLOE AIELLO @CHLOBO_ILO

Microsoft Says Workers are Already Using AI to Lighten Workloads, at Risk to Their Own Brainpower

Eccentric science-fiction author and technophile Douglas Adams once wrote about how tech was taking an effort-saving role in people’s lives: “Dishwashers washed tedious dishes for you, thus saving you the bother of washing them yourself,” he explained, and “video recorders watched tedious television for you” for much the same reason. But we’re in the AI era now, and a new Microsoft study suggests that Adams’s metaphor still applies: AI is able to take on much of that “tedious thinking” for you, saving you all the bother of actually working while at work. The new study actually warns that some knowledge workers are risking becoming overly reliant on generative AI, and their “problem-solving skills may decline as a result,” technology news site The Register says. The study acknowledges that people have objected to the impact of various technologies on the human mind since forever—from writing (a fundamental, ancient form of technology) all the way up to the internet. It also agrees that these worries are “not unfounded.” “Used improperly, technologies can and do result in the deterioration of cognitive faculties that ought to be preserved,” the authors write, noting that any type of automation can deprive people of chances to practice using their minds. In the survey, they asked 319 knowledge workers who used generative AI at least every week if they turn on their brains and apply “critical thinking” when they use tools like ChatGPT or Microsoft Copilot. The findings were stark. Survey respondents said that when they had high confidence that an AI tool would do well at a particular task, they felt less need to apply their own critical thinking. On the other hand, when a worker had high confidence in their own skills and less in the AI’s, they felt good about putting effort into evaluating the output the AI gave them and then improving it. AI is redefining how we see work It all boils down to the fact that when knowledge workers use AI tools, it shifts the way they think about performing activities like analysis, synthesis, and evaluation of information. The AI moves a worker’s focus from information gathering to information verification when using an AI to help try to understand something, and when using an AI for problem solving, the shift is away from carrying out the actual solving process to task stewardship. Think of it like this: When aircraft didn’t have an autopilot, fliers had to concentrate the whole time on operating the airplane—navigating, controlling, reacting to technical challenges, and feeling the way the wind was blowing. Modern day jetliner pilots have a very different job. They have to be able to fly the plane manually in case of unexpected problems, but minute to minute, what they’re often doing is monitoring the aircraft as it automatically flies itself to make sure it’s doing the right thing. Microsoft’s new research suggests that when people use an AI to help them solve work tasks, they’re doing the same thing—offloading the boring, slow, or difficult bits of the work to the AI and then managing the AI tool to get the desired output. The worry here is that over time people who used to hone their critical thinking skills all the time at work may lose some of that ability. One reassuring piece of pro-human info from the survey was that workers in high stakes workplaces or situations (like seeking medical advice from an AI) were conscious of the risk of over-relying on AI outputs that could be problematic, flawed, or flat-out wrong. Those respondents said they used their own thinking skills more. So what should we do about this? Should you worry that your workforce is going to become dimmer over time, human drudges merely shoveling data mindlessly into and out of an AI system? Not at all. The researchers suggest that one trick would be to design AI tools so they’ve got systems built into them that support worker skill development over the long term, The Register explains. And AIs should encourage workers to reflect on what’s happening when they’re interacting with AI outputs and even help the workers in this action—essentially keeping their minds focused, not blindly trusting the AI. It’s also possible that, as a good employer, you could give your staff tasks that keep their brains ticking over—ones that don’t need an AI boost. BY KIT EATON @KITEATON

Why do so many products that don’t seem to need AI integration still feel the need to include it?

The world didn’t ask for an AI-designed shoe. Nor did young parents across the country clamor for AI-enabled baby changing pads. It’s almost certain that your cat doesn’t know the difference between doing its business in a regular litter box versus one that is AI-enhanced. Yet all of these products exist, because the drumbeat of supposed progress in Silicon Valley is immutable. Super Bowl commercials were a prime example of AI’s societal stranglehold. OpenAI, Salesforce, and Google all aired commercials for their AI products during Sunday’s big game. How did we get here? We obviously didn’t start out with AI-powered teddy bears and a chatbot that specializes in erectile dysfunction. For the past couple of years, startups and tech giants alike have anchored themselves to AI hype because it’s been the easiest way to stay relevant or appear like the kind of mold-breaker that can turn the heads of almighty venture capitalists. After all, neither sovereign wealth funds nor institutional investors like stagnation. It’s been a trickle-down dynamic between AI’s influence on the tech world and consumer products in general. It started with OpenAI: Two years after initially getting bankrolled by Microsoft, the ChatGPT maker is closing a $40 billion investment from SoftBank, the Japanese investment titan. AI euphoria—or contagion—has swept Silicon Valley, and startups have the best chance of getting funding if they develop AI or otherwise integrate it into a product that’s apparently crying out for an overdue disruption—like the humble litter box. Last year, AI-related companies welcomed $100 billion in VC investment globally, which is roughly one-third of the $314 billion spent on all tech startups, a Crunchbase analysis shows. Undergirding all of this is a belief called “technological determinism,” Arun Sundararajan, a professor of technology, operations and statistics at New York University told Inc. last month. “If technology can do it, then it will happen,” Sundararajan explained. “As soon as the technological capability comes along, somehow, magically, it will enter our reality.” The latter half of that statement is undeniable. AI is being shoehorned into products that seemingly do not need to be enhanced by machine learning or large language models. Even in the summer of 2023, experts were warning that superfluous products riding the AI wave are eerily reminiscent of the firms that collapsed and caused market chaos in the dot-com crash. Last summer, Gayle Jennings-O’Byrne, CEO and general partner at VC firm Wocstar, said basically the same thing: “The mindset of VCs, versus the reality of what these business models and companies are going to look like, [is] just going to propel what we’re calling a bubble,” she told Inc. But inevitably, there will be more expansion, iteration, and pursuit of growth. Despite losing $5 billion last year, OpenAI is eyeing locations for new data centers in 16 different states, and Meta says it will allocate $65 billion for AI development this year. Even though ChatGPT has 200 million monthly users, it’s unclear to some experts whether the product adds anything of value to productivity and economic output. The worry is that consumer tools that can generate images and text on demand are more of a parlor trick than a new form of electricity. “We know ChatGPT has about 200 million unique monthly users, but the question is how many of them are using it in a way that will lead to significant productivity improvements/cost reductions. I don’t really know the answer to that question,” Daron Acemoglu, an economist at the Massachusetts Institute of Technology told NPR last October. BY SAM BLUM @SAMMBLUM

These Are the Jobs AI Will Replace

Question: Do you have a job that could be replaced by AI? OK, that was a trick question. Everyone’s job could be replaced by AI. That’s how they frightened CEOs into buying it. Better question: How do you know if your job will be replaced by AI? Let me answer that this way. There are two kinds of salespeople in this world. One kind is people who are good at selling and the other kind is people who are Salesforce Wizards. Similarly, there are two kinds of marketing people in this world. People who are good at marketing and HubSpot Gurus. You see where I’m going? To answer the question, we need to talk about the difference between expendable knowledge workers and irreplaceable knowledgeable workers. I’m Not Trying to Scare Anyone There’s actually a little bit of hope to talk about. For once. See, way back in the olden days—a.k.a., the summer of 2023—I predicted which jobs were most likely to be replaced by the coming Generative AI wave. TL;DR: I was one of the progenitors of early generative AI, in 2010, building a platform that enterprises like Yahoo and the Associated Press used to write insightful, informative narratives from nothing but raw data. Even back then, we knew that what we were doing would eliminate jobs. But everyone around us was confused as to which jobs our story-writing computers were going to eliminate. In 2010, we weren’t going to replace journalists or writers, at least not the good ones. Our tech was going to eliminate a new breed of “data scientists,” and only the sketchy ones. Those data scientists were knowledge workers. They knew how to use databases and SQL and R and Python to get insights out of the data. But it took the journalists, the knowledgeable workers, to make those insights make sense in context for the reader. I Was Right! Fast-forward to today and that battle is still going on. The threat has multiplied, of course, but not exponentially, because even today’s agentic AI is certainly not a font of unlimited contextual knowledge. What I learned back in 2010 and what still holds true today is that technical evolution has a way of calling out the rote-task knowledge workers in any industry. Back then, it was Johnny-come-lately data scientists. Today, it’s Salesforce Wizards and HubSpot Gurus. And AI does the calling out almost instantaneously, in a way that’s obvious when it’s not hallucinating. As I said in that 2023 article, it was only a matter of time before early-2023 generative AI was going to hit the knowledge economy, at which point, those rote-task knowledge workers should start worrying about their jobs. To clarify which jobs were in peril, I believe I used the phrase, “any white-collar, butt-in-a-seat, pixel-pushing, spreadsheet-spelunking job that the influx of data wrought on the workforce.” But even back in 2010, those jobs were starting to disappear, thanks to the automation that was and still is part and parcel of Big AI (or whatever). Those rote-task jobs were only being used as stepping stones to turn those knowledge workers into knowledgeable workers. In 2023, I said that most knowledge workers had about five to 10 years before they became obsolete. I Was Wrong! Well, we’re only in year three, maybe four, and corporate America seems hell-bent on eliminating the jobs of both knowledge workers and knowledgeable workers and letting God sort it out. That’s a huge problem. It has everything to do with how AI was sold into the enterprise, (i.e., FOMO), and that has been my problem with AI the whole time. In trying to reach maximum productivity, we just went all-in on the promise of AI and redefined productivity to meet it. However, as with most overreach cycles in business, I believe that’s finally changing. It might be too little and too late, but the AI bills are starting to come due. AI Hype Meets Financial Reality Even back in early 2024, obvious leaks in the AI productivity dream bucket started becoming very public. For example, this article from a fellow Inc. writer (go team!) digging into a survey from Upwork notes that 96 percent of C-suite executives expect AI to increase productivity, while 77 percent of employees actually using the tools as they exist today experienced decreased productivity. This isn’t a bright red warning flag or anything, but it does at least show the chasm-like mismatch in expectations versus reality that’s been snowballing over the last year. Now, in late 2023, I also said that AI was coming for SaaS, and everyone laughed at me again. Well, I was both right and wrong there too. I was right about AI replacing SaaS knowledge workers—those solely responsible for knowing how to get useful insights out of platforms like Google Analytics… or Salesforce or HubSpot. I was wrong in assuming corporate America would respond to this technical evolution sensibly and with caution and care for its employees. Boy, was I wrong. As I said, companies threw all kinds of babies out with all kinds of bathwater. To their own detriment. And here we are. We’re Not Out of the Woods Yet We’re not at the end of the AI hype cycle, but I believe we’re beyond peak AI hype. So the answer, the real answer for how to become irreplaceable, is the same and as simple as it ever was. Become a knowledgeable worker. Be unbeatable at what you do. Let AI handle the rote-task drudgery like staring at HubSpot all day. Because, yes, AI can come up with code or creative work or even make the hiring and firing decisions for you. It just can’t do it completely in context, and those skills which separate great coders and marketers and salespeople and CEOs are the same skills they always were. Look—it’s not how good you are at AI. It’s how good you are at everything that AI should not be doing. Which is a lot. OK. So now it’s just a matter of hiring back all those knowledgeable workers we lost—and are still losing—in the AI enterprise coup. Let’s hope our hiring system isn’t broken beyond all recognition. EXPERT OPINION BY JOE PROCOPIO, FOUNDER, JOEPROCOPIO.COM @JPROCO

The AI Energy Score attempts to quantify the environmental harm of AI models. Here’s how to use it.

Artificial Intelligence is notoriously energy-hungry. Now, a new tool from Salesforce is attempting to quantify the toll AI models are taking on the planet. “AI models vary dramatically in terms of their environmental impact, from big general models to small domain-specific models,” says Boris Gamazaychikov, head of AI sustainability at Salesforce. “What the AI Energy Score is aiming for is to develop a standardized way to actually start measuring these things.” AI Energy Score, launched Monday by Salesforce in collaboration with Hugging Face, Cohere, and Carnegie Mellon University, assigns a star rating to assess the energy efficiency of various AI models relative to other models. The ratings are available for 10 distinct tasks, from text and image generation to summarization and automatic speech recognition. A five-star rating is the most energy-efficient, whereas a one-star rating is least. It also shares a number estimate of the quantity of GPU energy consumed per 1,000 search queries in Watt hours. As of Feb. 10, DistilGPT2, a model developed by Hugging Face that is meant to be a “faster, lighter version of GPT-2,” was the most energy efficient, consuming 1.31 Wh of energy per 1,000 text generation queries. Alongside an interactive landing page with more information, the site also has a label generator that Gamazaychikov hopes developers will use to showcase their scores and drive awareness. For now, the tool only rates open-source models hosted on Hugging Face, a platform for data scientists, AI developers, and enthusiasts. That includes some models from Meta, Microsoft, Mistral, and StabilityAI. Salesforce also submitted its own open- and closed-source models for rating. Closed-source, or proprietary models, which include many big names such as OpenAI’s GPT-3 and 4, Anthropic’s Claude, and Google’s Gemini, are not readily available. But Gamazaychikov says there is a way for developers of these proprietary models to conduct the analysis securely and then submit it if they so choose. “By making a really easy to use, clear, standardized approach, we’re reducing any kind of friction between those leading companies being able to not do this,” Gamazaychikov says, adding that the tool was developed using feedback from academia, government officials, and AI companies. “We’re also hoping to increase some pressure, showing that it’s important to disclose this type of information.” Although Salesforce welcomes individuals to experiment with the tool to get a sense of how their use of AI models could take a toll on the environment, Gamazaychikov says “the primary audience is probably enterprises, or those that are integrating these types of models within their products.” Furthermore, a tool like AI Energy Score could help companies calculate indirect greenhouse gas emissions throughout the value chain. The launch of AI Energy Score follows Salesforce’s debut last fall of Agentforce, which allows customers to build or customize AI-powered agents that automate tasks or augment employee labor. Gamazaychikov says Salesforce also hopes to inspire greater change by giving regulators a jumping-off point for assessing the emissions-load of AI. He also envisions promoting the use of smaller, local models at a time when large reasoning models, which could consume even more energy than current large language models, are in development. A 2024 study from Goldman Sachs found that AI could drive up the energy demand of data centers by 160 percent by the end of the decade, even as the growing global population exerts additional pressures on the world’s energy needs. This comes at a time when leaders are contending with the worsening effects of climate change, prompting big tech companies to look to renewable energy and innovations in nuclear power to satisfy the growing demand. Want to try it out? Head on over to huggingface.co/AIEnergyScore and give it a whirl. BY CHLOE AIELLO @CHLOBO_ILO

Are you anxious about AI? Nvidia’s co-founder and CEO has some tips for how to find your feet in this new landscape.

If, during the past few years, you’ve felt like the future is barreling towards you like an oncoming train, you are not alone. The shock (and stock market dip) that accompanied last week’s announcement of a cheaper AI model from Chinese company DeepSeek shows that even experts can be blindsided by how fast tech innovation is happening. No one can take away that feeling of whiplash completely. Change is happening too fast for that. But if you’re looking for a good guide to help you get a handle on our AI-filled future, entrepreneur Jensen Huang should probably be at the top of your list. Huang is the CEO of Nvidia, maker of the chips driving the current AI boom. He built Nvidia into a $3-billion juggernaut by spotting the imminent rise of AI before just about anyone else and betting his company on it. Other tech CEOs fawn over his vision. What does Huang see for the future? Perhaps more importantly for entrepreneurs, how does he recommend you prepare? Huang: AI is like the interstate highway system That was the topic of conversation when Huang appeared on Cleo Abram’s Huge Conversations podcast recently. It’s an hour-long discussion, and if you’re fascinated by AI, then the whole thing is worth a watch. (I’ve embedded the complete interview at the end of this column.) Perhaps the most immediately actionable insight was Huang’s advice for everyday people wondering how best to prepare themselves for the coming AI revolution. On the podcast, Huang likens the change to the shift that arrived when the U.S. built the interstate highway system. Fast roads were the essential new technology at the heart of this change, but a whole ecosystem of other possibilities quickly developed around it. “Suburbs start to be created and distribution of goods from east to west is no longer a concern. All of a sudden, gas stations are cropping up on highways. Fast food restaurants show up. Motels show up, because people are traveling across the state,” Huang says. AI will be similar. Machines that can do many tasks better and faster than humans will create ripple effects that change many aspects of how we do our jobs and live our lives. How can you try to peek around the corner and get a glimpse of what that might look like? Huang suggests you ask yourself two key questions. If the drudgery it takes to do my job disappears, what changes? Some people worry that AI might take away jobs, making many workers superfluous. Huang doesn’t share this fear. He believes human insight and creativity will still be important, but what we spend our time on will be different. AI will kill rote donkey work. “Suppose that your job continues to be important, but the effort by which you do it went from a week long to almost instantaneous, that the effort and drudgery basically goes to zero. What are the implications of that?” Huang asks. Imagine you have an AI software programmer in your pocket that can write any software program you dream up. Or consider how it would impact your work if you could describe a rough idea and an AI could quickly produce a prototype for you to interact with. Innovations like these, Huang insists, shouldn’t make us feel threatened. They should make us feel empowered and excited about all the higher-level thinking and problem solving we’ll be freed to do. “I think it’s going to be incredibly fun,” he says. How can I use AI to do my job better now? If Huang’s first question is designed to get you thinking about what your work might look like 10 years from now, his second nudges you to consider what you can do now to prepare for that future. Huang tells Abrams that he has an AI tutor with him at all times. It’s a practice he recommends to just about everyone. “The knowledge of almost any particular field, the barriers to that understanding have been reduced,” he says. “If there’s one thing I would encourage everybody to do, it’s go get an AI tutor right away.” But don’t stop there. Huang’s more general point is that the more you experiment with AI now, the better prepared you’ll be to use it to your advantage as it develops. “If I were a student today, the first thing I would do is learn AI,” he declares a bit later in the podcast. He doesn’t mean learn technical details of the math behind the machines—though if you’re into that, certainly have at it. He means playing around with current tools like ChatGPT and Gemini to get comfortable with how to prompt them effectively. “Learning to interact with AI is not unlike being someone who is really good at asking questions,” he claims. “Promoting AI is very similar.” It’s a skill that requires honing. The end goal for everyone should be to begin thinking though how AI can best assist you with your work. “If I were a student today,” Huang continues, “doesn’t matter what field of science I am going to go into or what profession I am, I am going to ask myself, how am I going to use AI to do my job better?” Other AI experts agree with Huang Huang’s two questions are a great place to begin if you want to start to get a handle on how AI is going to affect you. But he’s hardly the only expert weighing in. There is no shortage of books on AI you can read to try and wrap your head around the technology. Other CEOs, like OpenAI’s Sam Altman and Bill Gates, have also weighed in on what our AI-future may look like and how to prepare. Even experts are still trying to figure out the future of AI, so don’t feel bad if you’re overwhelmed too. But while technologists are still building the future, they all agree we shouldn’t let anxiety or uncertainty get in the way of experimentation. The time for all of us to start thinking about the future of AI and playing with these tools is now. EXPERT OPINION BY JESSICA STILLMAN @ENTRYLEVELREBEL https://www.youtube.com/watch?v=7ARBJQn6QkM