Elon Musk is said to be moving forward with an artificial-intelligence project within Twitter, despite recently signing an open letter calling for an industrywide halt to any AI training for several months.The Tesla billionaire, who acquired Twitter almost six months ago and has made some drastic changes there, recently purchased roughly 10,000 graphics processing units for the platform, two people familiar with the company said.Tech companies typically use GPUs to work on large AI models, given the computational workload the newer technology requires.Musk’s AI project is at an early stage, according to the people familiar, though one person said the purchase of so much additional computational power showed he’s “committed” to its development.Musk, however, has criticized the recent development of generative AI, saying the technology is powerful and needs regulation to ensure it’s operating within “the public interest.”Musk’s AI project at Twitter involves a large language model, according to one of the people familiar.An LLM is essentially AI that is trained on large sets of data so it can eventually create new, seemingly sophisticated content and text on its own.Twitter has massive amounts of data that could train an LLM.For instance, ChatGPT’s creator, OpenAI, previously had access to Twitter data for training purposes, though Musk said he put a stop to that in December.Musk has also brought new AI talent to Twitter.By early March, he had officially hired the engineers Igor Babuschkin and Manuel Kroiss from Alphabet’s AI-research subsidiary, DeepMind.Since at least February, Musk has approached people in the AI field, including Babuschkin, about starting his own AI endeavor to rival ChatGPT, The Information reported.It’s unclear exactly what Twitter will use generative AI for, the people familiar said.
One use could be improving search, something Musk has complained about.He even hired the security developer George Hotz for a three-month internship to “fix” the search function.Hotz left after just one month.Another possible use is advertising, the backbone of social-media platforms, the people familiar said.Twitter has struggled to woo advertisers since Musk took over and changed numerous policies.
Generative AI has the capability, if trained for the use, to create new advertising images and text to target specific audiences.A Twitter representative didn’t respond to an email seeking comment.GPUs can be costly.Nvidia, which is estimated to have 95% of the market, manufactures a GPU for large AI models that costs $10,000.Musk, who has repeatedly said Twitter is on unstable financial footing, most likely spent many tens of millions of dollars to add so many GPUs, the people familiar said.The new units are expected to operate from one of Twitter’s two remaining data-center sites, most likely the one in Atlanta, the people familiar added.Twitter’s other remaining data-center site is in Oregon.Its main data-center site in Sacramento, operated by NTT, was abruptly shut down in late December in a unilateral decision by Musk.The engineering leader Behnam Rezaei resigned over the drastic decision.Musk has some history with ChatGPT.
He cofounded OpenAI alongside Sam Altman and other tech leaders.He walked away from the company in 2018 and has become critical of the company this year after it kicked off the ongoing generative-AI craze, saying it’s become “a closed source, maximum-profit company effectively controlled by Microsoft.” Microsoft invested in OpenAI in 2019 and in January said it was making a “multibillion dollar investment” as part of an extended partnership.Are you a Twitter employee or someone else with insight to share? Contact Kali Hays at [email protected], on the secure messaging app Signal at 949-280-0267, or through Twitter DM at @hayskali.Reach out using a nonwork device..
