Microsoft and OpenAI have parted ways, and are now ready to fight

At Microsoft’s annual Build conference on Tuesday, the company announced a slew of new or expanded AI initiatives, including a super app, internal inference models, a cybersecurity tool, and OpenClaw-esque AI agents. All of this news adds up to a clear message: Microsoft is positioned to be one of the biggest players in AI, and it’s finally acting like it.

For many years, Microsoft’s AI business has relied heavily on its early and exclusive partnership with OpenAI. But the drama-filled marriage slowly developed into a situation and a husband Effectively separate In late April (although Microsoft is still OpenAI’s primary cloud partner – for now). This year’s design had the vibe of a modern single divorcee posting a thirst trap on Instagram. “It’s always fun to be at developer conferences in times of great change,” Microsoft CEO Satya Nadella said on stage Tuesday, adding that such events are about “dealing with new opportunity.”

The head of Amnesty International, Mustafa Suleiman, in an interview with EdgeIn other words, more frankly.

“The goal is to prove that we can become one of the top four laboratories in the world,” Soliman said. “There are three important labs, Google DeepMind, OpenAI, and Anthropic. We’re not one of them right now, and that was always my intention. That’s why I came here. I want to build the best frontier models in the world, fully multimodal, and in order to do that, we have to prove that we can do everything we need from A to Z, and not just take from others.”

One of Microsoft’s first steps in Build was to catch up with AI models. Solomon unveiled MAI-Thinking-1, the company’s first thinking model, along with six other new models focusing on image, audio, transcription and encryption. Microsoft He said The mid-sized MAI-Thinking-1 model, which will likely be marketed primarily to enterprise customers, is “designed from the ground up for serious mathematics, programming, and real-world enterprise deployment.” Microsoft is years behind both OpenAI and Anthropic here; OpenAI will begin releasing inference models in the fall of 2024. But Soliman emphasized their performance on criteria like programming and price point, saying they are cheaper than their OpenAI counterparts on some tasks, which is a big deal in the age of AI. AI money squeezeWhich inspired a lot of complaints with customers.

While Microsoft has had years to extract insights from OpenAI, Solomon is careful to point out that its development did not involve any distillation, meaning it was not trained using a different company’s AI model. If MAI-Thinking-1 is good, Microsoft clearly doesn’t want people to think it’s due to the influence of OpenAI.

Solomon said Edge For Microsoft, “the pivotal moment was renegotiating our contract with OpenAI. This meant that we were allowed to train models at a larger scale and go after superintelligence entirely explicitly using our own IP, with our own data, without distillation, and training from scratch.”

Nadella also highlighted Microsoft’s recently launched AI cybersecurity tool MDASH, which he said brings together 100 AI agents to find exploitable bugs “better than any single model.” Apparently it was research in the Claude Mythos Preview, which Anthropic presented in April A lot of fear and fuss -And Expand your reach Immediately before construction. OpenAI has Its own system focused on cybersecurity Also, the three companies are likely to use their bids to compete for positions in the government and institutional markets that desperately need them on the court.

Microsoft faces a more complicated situation with AI agents. the Popular open source platform OpenClaw demonstrated a potential path forward for AI clients, and then OpenAI was quickly employed Its creator, Peter Steinberger, Microsoft (Among other companies) is trying to catch up. One of its main strategies is to make OpenClaw work well with Windows. At Build, Nadella said he was very committed to supporting OpenClaw, and Microsoft employees spoke with developers in the audience about how they use it.

Steinberger himself made a surprise appearance to great reaction from the audience, taking to the stage to show off how OpenClaw has strengthened its security and earned user trust. “What I kept hearing was, ‘Peter, I love my Claw, but can I use it at work?’” Steinberger said. “You can run OpenClaw completely in-house now, and we’ve even made the tool itself a plug-in.” Whether someone trusts Copilot, Codex, or another company’s programming platform, users can now run OpenClaw on top of that via Windows, Steinberger said.

But Microsoft is also promoting its own separate Copilot super app that integrates OpenClaw-esque proxies. The super app is a major focal point for OpenAI right now – with President Greg Brockman leading the development of an app that will connect ChatGPT, the Codex programming platform, and the Atlas web browser. Microsoft’s strategy is similar, combining a variety of existing Copilot AI assistants. Its agents, called “autopilots,” are designed to serve as a useful user interface. Cassidy Williams, senior director of developer support at GitHub, called Copilot “your home base for development and operations on your computer,” explaining how multiple agents can perform tasks like building applications. (In added brilliance, Williams demonstrated how she can approve or reject code changes by raising a thumbs-up or thumbs-down sign on her computer camera.)

Autopilots are specifically designed to appeal to business clients – Nadella called them “autonomous agents that operate for the long haul with full enterprise compliance.” The first thing Microsoft will offer is “Scout”. described as “Your own permanent agent,” but customers can build and customize their own agent. Autopilot agents should be able to check their email inbox, join group chats in Teams, check their calendar, and send daily summaries, among other things. Accordingly, employees on stage at Build repeatedly emphasized Copilot’s security tools and guardrails — clearly intended to placate enterprise customers who might have overheard Horror stories About tools like OpenClaw.

Sulaiman has been careful to repeatedly emphasize Microsoft’s “human superintelligence” as “AI that puts humanity first” — part of the company’s latest AI initiatives. Rebranding of AGI To make it seem less scary in an age where people are Rollback against artificial intelligence More than ever.

Nvidia CEO Jensen Huang, another speaker known for working closely with OpenAI, appeared via video call to promote how Nvidia’s RTX Spark chip advances Microsoft’s AI agent goals. “I might be traveling and I’m on the phone and I can send a text message to my computer… and that will trigger the tools on the computer,” Huang said. “The idea of ​​a personal computer evolving from a personal computer to a personal AI is really exciting.”

Microsoft has spent years betting on OpenAI and, in some ways, leaving it behind in the AI ​​race. But as OpenAI and other competitors eventually turn to enterprises to make money, it has some clear advantages. Microsoft already has a large customer base and has a good reputation for safety and security compared to other AI companies. and Like GoogleThey also have significant resources, significant computing resources, and a diversified revenue stream, which means they can take large bets without much risk.

Solomon said Edge“There are a lot of people who either want to chase startup valuations or are about to go public, so we can operate with more humility and less optimization in the long term.” “We have the money to be able to buy Anthropy (models) when we need to. We have an option in Azure with 11,000 models, so people can literally use whatever they want whenever they want, but it gives us the time to do it right from the beginning,” he added.

At the same time, there are a lot of unanswered questions here. Microsoft has had a lot of wins and benchmark advancements for its seven new models, but that doesn’t always translate into real-world adoption, or even a new model. Which progresses for a week or two It can fall behind quickly. AI superapplications are mostly an untested idea. And Microsoft is entering a crowded but still largely disappointing market for AI agents with a product we haven’t seen in action. There is still plenty of room for its promises to fail.