At the end of March, Elon Musk joined a group of more than 1,000 academics and engineers calling for a six-month pause in the development of artificial intelligence tools, such as GPT-4. They consider that the dangers of artificial intelligence for society are too great not to define at least some ethical guidelines or establish certain regulations.
But despite signing this petition, Musk is actively working on creating a language model similar to those of OpenAI or Google. He is doing it in secret, with a select group of developers on Twitter and with the goal of offering an alternative in the coming months.
According to Business Insider, Musk would have recently purchased about 10,000 graphics cards to develop and train this model, which will be installed in one of Twitter’s data centers. Since the Twitter purchase, Musk has embarked on drastic cost cutting that has led him to lay off the vast majority of the company’s employees and close several of its data centers, but he still maintains two large facilities in the US. , in the cities of Atlanta, Georgia and Portland, Oregon.
Most AI applications benefit greatly from this type of component because graphics cards, unlike processors, are designed to perform enormous amounts of computations in parallel, and these types of computations are precisely what is required during gaming. training process and execution of language models.
The models that are used, however, are different from the graphics cards that are usually associated with desktop PCs and video games. These are components specifically created for calculation tasks and that can cost tens of thousands of euros.
Musk’s project started in February and is still in its initial phase, but the magnate has hired several experts and engineers from DeepMind, an artificial intelligence company that is a subsidiary of Alphabet, the Google business group, to lead the team.
Musk was one of the initial promoters of the OpenAI project, creators of the GPT-4 language model and the famous ChatGPT tool, but he left the organization in 2018 after failing to become the visible head of the project.
Gaining board support, Sam Altman has managed to transform the organization from a nonprofit to one of Silicon Valley’s most admired companies. In January of this year, Microsoft announced an investment of close to 10,000 million dollars in the company in exchange for 49% of the shares and 75% of the profits generated in the coming years until the amount contributed is recovered.
According to the criteria of The Trust Project