Microsoft is bringing the Grok 3 and Grok 3 mini AI models from Elon Musk’s xAI to its Azure AI Foundry platform, as new offerings hosted and billed directly by Microsoft from within the company’s cloud platform.
The move is notable in part because it shows Microsoft and xAI working more closely together, even though Elon Musk is publicly at odds with OpenAI, Microsoft’s biggest AI partner. Musk last year added Microsoft as a defendant to his lawsuit against OpenAI, alleging that the companies formed a de facto AI monopoly.
Azure AI Foundry is a platform within Microsoft Azure designed to help developers build, customize, and manage AI applications and agents. It includes models such as OpenAI’s GPT-4, Meta’s Llama 3, Mistral models, and Microsoft’s own small language models (SLMs), including Phi-3.
Musk appeared with Microsoft CEO Satya Nadella in pre-recorded video highlights shown at the event, and Microsoft made the full video available on YouTube, which you can watch above.
If there was any tension between Musk and Nadella, it wasn’t evident in the recorded video.
- Nadella started the conversation by referencing Musk’s early ties to Microsoft, mentioning his time as a Windows developer and Microsoft intern, and noting his continued enthusiasm for PC gaming.
- Musk looked back even further, reminiscing about programming games in DOS before moving on to Windows.
- Nadella also recalled previous chats with Musk about technical topics like Active Directory.
More highlights:
On Grok’s core architecture: Musk stated that Grok is “trying to reason from first principles.” He explained this means seeking “fundamental truths,” boiling “things down to the axiomatic elements that are most likely to be correct, and then you reason up from there.”
He likened the approach to applying “the tools of physics to thinking” across all lines of reasoning, with the goal of arriving at “truth with minimal error.”
On AI safety: AI safety is a key concern for Musk, who emphasized that it’s closely tied to transparency and integrity. His conclusion: “Honesty is the best policy. It really, really is for safety.”
While acknowledging that “we have and will make mistakes,” he stressed the importance of quick corrections and ongoing input: “We are very much looking forward to feedback from the developer community.”
On grounding AI in reality: A core principle behind Grok, Musk said, is that intelligence must be anchored in the physical world. “Physics is the law, and everything else is a recommendation.”
He added, “I’ve seen many people break human-made laws, but I have not seen anyone break the laws of physics.” This grounding, he argued, is critical to ensuring that AI models are truthful and accurate.
On real-world testing and deployment: Musk highlighted how real-world use cases validate Grok’s performance.
“The car needs to drive safely and correctly. The humanoid robot Optimus needs to perform the task that it’s being asked to perform. These are things that are very helpful for ensuring that the model is truthful and accurate.”
He noted Grok is already proving “very helpful in things like customer service,” describing it as “infinitely patient and friendly… you can yell at it, and it’s still going to be very nice.”
On current performance and future outlook: Grok is “already doing quite a good job at SpaceX and Tesla,” particularly in customer support roles, Musk said.
Looking ahead, he expressed a desire to broaden its reach: “We look forward to offering that to other companies.” He closed with a message to developers: “Tell us what you want, and we’ll make it happen.”
Earlier in the event, OpenAI CEO Sam Altman joined Nadella via live video. They talked about how AI is changing software development, and discussed how AI tools are becoming more like helpful teammates that can take on complex coding tasks. Altman said AI models are improving quickly, getting better at understanding different types of input and working more smoothly with other tools.
Microsoft and OpenAI are reportedly renegotiating their multibillion-dollar partnership, according to the Financial Times. The talks come as OpenAI moves forward with a corporate restructuring that would shift its for-profit operations into a new public benefit corporation, still controlled by its nonprofit parent.
Read the full article here
