Xai co-founder Igor Babuschkin at the NVIDIA GPU Technology Conference (GTC) held in San Jose, California, USA on Tuesday, March 19, 2024.
David Paul Morris | Bloomberg | Getty Images
Igor Babuschkin, a founding member of Elon Musk's Xai, said on Wednesday he will leave the artificial intelligence startup and launch his own venture.
“Today was my last day at Xai, helping me get started with Elon Musk in 2023,” Babuschkin wrote to X, owned by Xai. “I still remember the day I met Elon. We spent hours talking about AI and the future could happen. We both felt that we needed a new AI company with a different kind of mission.
Musk wrote, “Thank you for helping us build @xai! We wouldn't be here without you.”
Babuschkin said he is launching the Babuschkin venture to support AI safety research and invest in startups with “AI and agent systems that advance humanity and unleash the mysteries of our universe.”
Former research engineer of Google's Babuschkin spoke about some of Xai's key operational results during his tenure, Deepmind, technical staff and former member of Openai.
“Through the sweat and tears of blood, our team's furious speeds have built a Memphis super cluster and shipped frontier models faster than any company in history,” he writes.
The Memphis facility will work with data and train models that power Xai's Grok Chatbot.
Locals are protesting Xai's operations in Memphis, particularly using natural gas combustion turbines to strengthen their data centers. Turbine emissions are reportedly exacerbating the decline in air quality in West Tennessee.
As he was preparing to enter business with Musk, Babushkin wrote that he believed that “AI can infer immediately beyond human levels” and was worried about making sure such technology would be “used for good.”
He shared his vision of “AI used to benefit humanity,” saying, “Eron had been warning of the dangers of powerful AI for many years.”
The company has a rocky track record when it comes to AI safety.
In May, Xai's Grok Chatbot automatically generated and spread false posts about allegedly “white genocide” in South Africa. The company then apologised, saying Grok's strange behavior was caused by a “fraudulent change” to Chatbot's system prompt.
In July, Xai apologised for another issue with Grok. After updating the code, the chatbot will be automatically generated and spread with false anti-Semitic content throughout X, including posts praising Adolf Hitler.
The European Union last month called for a meeting with Xai representatives to discuss the issue with X-integrated Grok chatbots.
Babuschkin and Xai did not respond to requests for comment.
Other chatbots also generated erroneous or dangerous output depending on the query. Openai's ChatGpt was recently called to give bad health advice to users caught up in emergency rooms. Google also had to make changes to Gemini last year after generating aggressive images in response to user prompts about history, including images depicting people of Nazi colour.
Watch: Xai explains Grok's “White Genocide” post