Long before OpenAI’s ChatGPT took the internet by storm, Elon Musk was a regular AI critic. But last month the owner of Twitter went as far as to sign an open letter calling for a pause on further AI training due to potential dangers with artificial intelligence.
But, while Musk has called for an AI pause, he apparently is going ahead with an AI project of his own.
According to a report from Insider, Elon Musk has recently acquired 10,000 GPUs, an investment for a new AI project at Twitter. It’s unclear exactly what the AI project is. However, Insider’s sources said that Twitter is planning to use its data to train a large language model or LLM and speculated that it could be for updating Twitter’s search or ad serving tools.
OpenAI has used Twitter’s data to train its own language models. However, Musk said in December that he was “pausing” OpenAI’s access to Twitter’s database.
Working on large AI models is an intensive, and costly, process. Insider points out that GPUs for AI training made by leading manufacturer Nvidia can cost roughly $10,000, putting Musk’s purchase at a likely price tag in the tens of millions of dollars.
Musk has been unhappy with OpenAI, which he co-founded in 2015 and has since cut ties with the organization. Since ChatGPT became the fastest growing app of all time earlier this year, Musk has hinted at creating his own “based” ChatGPT competitor. He’s also taken shots at Microsoft, OpenAI’s largest partner, which has integrated OpenAI’s language models into its Bing search engine. Last month, Musk questioned Microsoft’s decision to cut its AI ethics team, even though he laid off Twitter’s very own AI ethics team when he took over the company.
At Tesla, Musk has been much less gloomy about certain kinds of AI, thanks to the electric car company’s need for these technologies in the development of self-driving vehicles. Time will tell if Musk’s opinion of AI chatbots shifts as well once he’s integrated it into one of his own business models.