OpenAI Researchers Leave to Build Next-Generation AI Startups

Several leading researchers have left OpenAI in recent years to launch new artificial intelligence ventures. These departures reflect a broader trend of top talent seeking to develop advanced AI technologies outside the company.

Key Highlights

• Several OpenAI researchers have left to start new AI ventures in recent years.
• Naomi Bashkansky joined Conduit to develop thought-to-text AI using non-invasive brain signals.
• Former OpenAI chief scientist Ilya Sutskever launched Safe Superintelligence with a focus on safe AI.
• Nvidia reportedly invested $5 billion in Safe Superintelligence for early access to new AI chips.

Recent Departures from OpenAI

Naomi Bashkansky is the latest researcher to leave OpenAI. She announced her resignation in late July and joined Conduit as a founding researcher the next day. At Conduit, Bashkansky will work on AI models that convert human thoughts into text using non-invasive brain signals. She shared her reasons for leaving in a blog post titled "Why I'm leaving OpenAI to build telepathy." Bashkansky explained that Conduit's goal is to develop AI systems capable of translating a person's thoughts into text without surgical brain implants.

Conduit is building AI models that interpret brain signals collected through non-invasive technology. Bashkansky outlined a roadmap for this technology, projecting advancements by 2027, 2030, and 2035. The long-term aim is to enable people to communicate by thinking, without the need for physical devices or invasive procedures.

New Startups from Former OpenAI Leaders

OpenAI has seen several high-profile exits as researchers and executives pursue new AI ventures. One of the most notable departures is that of OpenAI co-founder and former chief scientist Ilya Sutskever. He left the company in 2024 after a period of internal leadership changes. Sutskever then launched Safe Superintelligence (SSI) with Daniel Gross and Daniel Levy.

SSI has a different focus compared to companies like OpenAI, Google DeepMind, and Anthropic. Instead of developing AI chatbots or consumer products, SSI aims to build "safe superintelligence." This means creating AI that surpasses human intelligence while prioritizing safety in its development. SSI has raised billions of dollars from investors but has released little information about its technology.

Recently, SSI has attracted attention after venture capitalist Gavin Baker suggested on the Invest Like the Best podcast that SSI could unveil its first AI model this month. The company has not confirmed any launch date or provided details about the model's capabilities. Industry experts believe the new model may demonstrate research conducted by Sutskever and his team in private.

Reports indicate that Nvidia has invested $5 billion in SSI, giving the startup early access to next-generation Vera Rubin AI chips. These chips are expected to support SSI's research and development efforts. Despite speculation, SSI has not officially announced plans to release a model in August.

Other Industry Developments

Anthropic is reportedly working to develop its own chips to power its Claude AI system and is building a dedicated team for this purpose. In other news, Microsoft has retired its employee feedback tool and now encourages staff to communicate face to face. Zoho's CEO commented that flexible work arrangements have existed at the company for 30 years, predating Gen Z's influence.