Elon Musk has started a project within Twitter to train his own AI. To this end, the company has already purchased 10,000 GPUs and hired new talent.

Do what I say, not what I do: This is probably the mantra of Twitter CEO Elon Musk. He recently joined international experts in a call to put AI development on hold. Musk himself has no intention of hitting the pause button. On his impulse, Twitter would have bought about 10,000 GPUs to train its own AI.

The project is still in its infancy. With the GPUs, Twitter at least has hardware to become one with large language model how GPT-4 can train. To steer the project in the right direction, Twitter would have already hired talent from DeepMind.

Much ambiguity

It is unclear which GPUs are involved. In any case, the investment would be worth tens of millions. This is noticeable because Twitter is currently not doing well financially. Where the GPUs will end up is also vague. After all, Musk closed Twitter’s main data center in Sacramento in the US last year. The company still has a suitable location in Atlanta.

Musk and Twitter therefore want to build their own AI, but do not reveal why. Musk himself wanted to step down from taking over Twitter at the time because he felt there were too many bots on the platform. The future will show to what extent an extra smart AI bot is a good idea.