xAI co-founder Igor Babuschkin leaves to start AI safety fund

In short: Igor Babuschkin, a co-founder of xAI, is leaving the company to start Babuschkin Ventures, a new firm aimed at funding AI safety work and startups.

What happened

Igor Babuschkin helped found xAI with Elon Musk in 2023. In an interview, he said he is leaving to launch Babuschkin Ventures.

Babuschkin Ventures will focus on funding AI safety research and startups. AI safety research is work meant to reduce risks from AI systems, such as systems that can be misused or behave in unexpected ways.

Babuschkin also said he wants people to be able to customize AI models for their own advantage and have more control over their future. In simple terms, that means more people could adjust how an AI works to fit their needs, instead of relying on one big company’s version.

Why it matters

This connects to a wider debate about how open AI should be. xAI has released some of its AI model files publicly, including Grok-1 in March 2024.

When a company releases a model’s “weights” and “architecture,” it is like sharing the finished engine and the blueprint so others can inspect it and tune it. xAI released Grok-1 under the Apache 2.0 license, which generally allows broad reuse.

However, the release did not include the training data, which is the large collection of text and other information used to teach the model. So people can study and modify the model, but they cannot fully see what it learned from.

Elon Musk has also said xAI would keep opening up its models. He later said Grok 2.5 was open sourced, and that Grok 3 was planned for an open-source release about six months later.

Source: NYTimes