Anthropic's Claude 4 Opus is cutting edge AI
If Anthropic succeeds
Microsoft’s and Google’s opposite AI strategies
Anthropic's Claude 4 Opus is cutting edge AI
The new models, which jump the naming convention from 3.7 straight to 4, have a number of strengths, including their ability to reason, plan, and remember the context of conversations over extended periods of time, the company says. Claude 4 Opus is also even better at playing Pokémon than its predecessor. Hershey’s overarching goal with this research(playing Pokémon) was to study how Claude could be used as an agent—working independently to do complex tasks on behalf of a user.
Anthropic, like many other AI labs, is hoping to create powerful agents to sell as a product for consumers. Krieger says that Anthropic’s “top objective” this year is Claude “doing hours of work for you.” "This model is now delivering on it—we saw one of our early access customers have the model go off for seven hours and do a big refactor,” Krieger says, referring to the process of restructuring a large amount of code, often to make it more efficient and organized.
This is the future that companies like Google and OpenAI are working toward. Earlier this week, Google released Mariner, an AI agent built into Chrome that can do tasks like buy groceries (for $249.99 per month). OpenAI recently released a coding agent, and a few months back it launched Operator, an agent that can browse the web on a user’s behalf. Compared to its competitors, Anthropic is often seen as the more cautious mover, going fast on research but slower on deployment….
***
If Anthropic succeeds
The brother goes on vision quests. The sister is a former English major. Together, they defected from OpenAI, started Anthropic, and built (they say) AI’s most upstanding citizen, Claude. And that's why businesses like Claude - “brand safe”. They also seem to be on the cutting edge too. Amodei(CEO of Anthropic) has just gotten back from Davos(where billionaires tell millionaires how the people are feeling), where he fanned the flames at fireside chats by declaring that in two or so years Claude and its peers will surpass people in every cognitive task.
Back when he worked at OpenAI, Amodei wrote an internal paper on something he’d mulled for years: a hypothesis called the Big Blob of Compute. AI architects knew, of course, that the more data you had, the more powerful your models could be. Amodei proposed that that information could be more raw than they assumed; if they fed megatons of the stuff to their models, they could hasten the arrival of powerful AI. The theory is now standard practice, and it’s the reason why the leading models are so expensive to build. Only a few deep-pocketed companies could compete.
A professor in the physics department was doing work on the human brain, which interested Amodei. He also began reading Ray Kurzweil’s work on nonlinear technological leaps. Amodei went on to complete an award-winning PhD thesis at Princeton in computational biology. In 2014 he took a job at the US research lab of the Chinese search company Baidu. Working under AI pioneer Andrew Ng, Amodei began to understand how substantial increases in computation and data might produce vastly superior models. Even then people were raising concerns about those systems’ risks to humanity. Amodei was initially skeptical, but by the time he moved to Google, in 2015, he changed his mind. “Before, I was like, we’re not building those systems, so what can we really do?” he says. “But now we’re building the systems.”
Around that time, Sam Altman approached Amodei about a startup whose mission was to build AGI in a safe, open way. Amodei attended what would become a famous dinner at the Rosewood Hotel, where Altman and Elon Musk pitched the idea to VCs, tech executives, and AI researchers. “I wasn’t swayed,” Amodei says. “I was anti-swayed. The goals weren’t clear to me. It felt like it was more about celebrity tech investors and entrepreneurs than AI researchers.” Months later, OpenAI organized as a nonprofit company with the stated goal of advancing AI such that it is “most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.” Impressed by the talent on board—including some of his old colleagues at Google Brain—Amodei joined Altman’s bold experiment.
At OpenAI, Amodei refined his ideas. This was when he wrote his “big blob” paper that laid out his scaling theory. The implications seemed scarier than ever. “My first thought,” he says, “was, oh my God, could systems that are smarter than humans figure out how to destabilize the nuclear deterrent?” Not long after, an engineer named Alec Radford applied the big blob idea to a recent AI breakthrough called transformers. GPT-1 was born….
***
Microsoft’s and Google’s opposite AI strategies
Microsoft’s Build and Google’s I/O conferences were all about AI, the dueling convocations highlighted how the two industry behemoths are each seeking to conquer the market through radically different strategies. But beyond coding agents, some key differences in emphasis pointed at divergent strategies.
Microsoft is battling to convince enterprises to build AI agents.
At Build, Microsoft placed a far greater emphasis in its announcements on tools that are designed to help enterprise customers create AI agents and get them to successfully automate workflows. Microsoft’s announcements were about how to allow agents to use tools, get agents to work with other agents, and, critically, to control what data AI agents access. These things matter to big companies and governments.
Google is battling for consumers and individual creators
Contrast that to what Google announced at I/O. Here the emphasis was almost entirely on consumers, not large organizations. It was about individual web users and individual content creators. The biggest news was the revamping of Google’s core Search product, with more AI Overviews, which provide capsule answers to queries, and also a new “AI Mode” that provides a more native AI experience, similar to what users get with OpenAI’s ChatGPT, using Google’s most capable AI models. It will also have new features that allow shoppers to virtually try on outfits as they shop. Sure, there was some talk about agentic AI capabilities—which are being released under what Google is calling Project Mariner—but these were about agents designed to help consumers do things, like purchase tickets to a sporting event, or buy groceries. Project Mariner is about building a universal personal assistant. It is not about automating enterprise workflows.
Microsoft needs the tech to work. Google needs that—and a new business model.
What was also striking between Build and I/O is how comfortably the innovations Microsoft is announcing sit within the software giant’s existing business model, and how awkwardly much of what Google announced sits within its own. Sure, Microsoft is taking a risk that its customers won’t find enough value in all the agentic AI products and features it is rolling out to pay the increased license fee that Microsoft wants to charge for it. But, if the AI agents gain traction, they only reinforce its existing cloud business and subscription-based business model.
Google, on the other hand, is taking a big gamble with its rollout of AI features that could directly cannibalize the advertising-based business model on which it has depended for a quarter century. Search represents 56% of Google’s revenues and most of its profits. If people click on fewer links with AI Overviews, as independent studies suggest, or if AI Mode offers far fewer opportunities for paid links, as also seems to be the case, it isn’t clear how Google will maintain its revenues. There are plenty of ways to imagine new business models for chatbot-like interface and a universal personal AI assistant. But Google has not said yet what it thinks those business models should be—and listening to Google executives speak at I/O one got the sense the company hasn’t really figured it out yet….