Since OpenAI board members ousted him on Friday, November 18, Sam Altman’s next move has been the subject of intense speculation. Will he mount a comeback as the Artificial Intelligence (AI) firm’s CEO? Launch a new chipmaker to rival NVIDIA? Or perhaps pursue his AI hardware project with Apple’s Jony Ive?
Altman is hardly short of options when it comes to his next steps. Anywhere he ends up, employers will have to deal with potential conflicts of interest arising from the tech founder’s vast entrepreneurial ambition.
Officially, Altman was fired from OpenAI because he was “not consistently candid in his communications with the board.”
Before the dust had even settled on his unexpected departure from OpenAI, on Monday morning, Bloomberg reported that Altman had spent the previous weeks touring the Middle East, soliciting investment for a new venture: Tigris.
According to the report , the Tigris project sought billions of dollars to found a new chip manufacturer geared toward the AI sector.
In a semiconductor market dominated by NVIDIA, the proposed chipmaker would research and develop Tensor Processing Units (TPUs) – a type of application-specific integrated circuit (ASIC) designed to handle machine learning workloads.
But Sam Altman’s interests extend far beyond AI training hardware.
A second potential distraction that may have complicated his position at OpenAI is Altman’s collaboration with Apple’s former chief design officer, Jony Ive.
According to a recent Financial Times report, Altman tapped Ive’s design firm LoveFrom to help develop a new consumer device for OpenAI.
With Softbank CEO, Masayoshi Son, reportedly prepared to bankroll the project, the two Silicon Valley bigshots are said to have been inspired by the way the iPhone’s touchscreen transformed the way people use the internet. However, with no deal signed and OpenAI’s charismatic CEO now freelance, there’s no guarantee the company will have a role in any future development.
Between TPU’s and the iPhone for AI, Altman’s interest in hardware sends a clear message.
Until now, AI has borrowed or repurposed technology originally designed for different purposes. But as the space evolves, custom-built equipment could come to play a much larger role.
In a recent interview with CCN, Rajiv Khemani, the CEO of Auradine, a company that builds Bitcoin mining equipment, discussed the limitations of the technology that companies like OpenAI currently use in their AI-training supercomputers.
“I think things will need to become more efficient and secure going forward because the current trajectory is not sustainable in the deployment of these leading-edge GPUs,” Rajiv Khemani observed.
Of course, the field of Bitcoin mining had its own ASIC revolution when Auradine’s predecessors launched the first purpose-built mining rigs in the previous decade. Meanwhile, in the field of AI, dedicated hardware is only just emerging.
Google started using cloud TPUs to train machine learning models in 2015 but only created the first custom AI chip in 2018. With the technology still in its infancy, the potential rewards are huge for any company that can further optimize TPUs and bring the first widely accessible, scalable product to market.
Meanwhile, consumer-facing AI interfaces promise to be just as lucrative in the coming years.
According to one study , the market for smart speakers like Amazon Alexa and Google Home is expected to reach $100B by 2032. But for the kind of device that Altman and Ive are reportedly working on, the $500B+ smartphone industry may be an even better benchmark.
For now, Altman has found employment with Microsoft, where he is heading up a new advanced AI research team alongside fellow OpenAI exile Greg Brockman.
In the end, whether it’s at Microsoft, Tigris, or some other as-yet-unidentified venture, as the AI sector continues to evolve, I’m sure there’s more to come from one of its most restless innovators.