Listen to this story
Elon Musk loves media attention. This time it’s about AI and his all evil brainchild ‘X Corp’. The most-celebrated critic of OpenAI’s trending technology, who calls for a pause on training anything beyond GPT-4, is now building his own AI project.
According to reports from Insider, Musk has bought thousands of GPUs in a bid to get ahead in the generative AI race. This comes just over a month after he hired two of DeepMind’s researchers, Igor Babuschkin and Manuel Kroiss, for building a rival to OpenAI “woke” ChatGPT.
It seems like the one who is considered an AI-doomer for signing the petition to put a pause on training AI models beyond GPT-4, was actually just in it for his self-interest. Though it is unclear if the two researchers he hired are working for Twitter or not, it is definitely clear that they are working for Musk.
Twitter Data is a Gold Mine
Whatever Musk plans to do in generative AI, he has one thing that no other company, even OpenAI, has — Twitter’s data. So, if the reports are true that Musk is indeed working on a project that involves LLMs, then the data that he has access to is a gold mine. Earlier OpenAI had access to Twitter’s data, which Musk pulled away after buying the company. Seems like this was a plan all along.
When it comes to Twitter as a company — it does not exist anymore. Musk has merged Twitter with his newly formed shell firm called X Corp. This was revealed in the document of a lawsuit filed against Jack Dorsey, the former CEO of Twitter, by Laura Loomer.
Musk’s long ambition of building a superapp, something similar to China’s WeChat, seems to be now taking place in the form of X Corp, an “everything app”. To mark the start of this move and the place where Twitter ceased to exist, Musk Tweeted “X” on Tuesday. He had already suggested that buying the social media company was just an “accelerant” for creating X.
Today, Musk replied to a BBC journalist on Twitter Spaces about how X Corp is definitely in the making. Previously, Musk had also tried to name PayPal as X.com, which received major backlash from the board members, eventually leading to him dropping the idea.
Musk has been trying to fix Twitter for a while now on different fronts. He had complained about the search function on the platform, for which he hired George Hotz. Hotz eventually quit within one month of the three-month internship saying that he felt he “didn’t think there was any real impact“.
Another thing that Twitter has been trying to win is advertisers’ interests. Companies have been criticising the change in policies after Musk took over. According to sources, if generative AI is used for the purpose of advertising, it can be beneficial for Twitter. Interestingly, when Musk recently open-sourced Twitter’s algorithm, he did not disclose the ad recommendation system.
Musk’s Epic Fails
Musk is a ‘walking-talking’ sinking ship. He had previously expressed his intention of turning Twitter from a loss-making to profit-generating company. For this, after taking over Twitter, Musk had been on a firing spree. He laid off 3,700 people in November, which is almost half the staff. He had also shut down the India-arm of the company, firing over 200 employees.
Interestingly, a single NVIDIA GPU used for large AI models costs $10,000. If the company had to cut down employees citing unstable financial footing, funding millions of dollars for buying thousands GPUs for generative AI seems to be a messed up move. In today’s Twitter Spaces discussion, Musk expressed the importance of buying GPUs.
Meanwhile Tesla’s supercomputers are only being used for Tesla’s Autopilot and Optimus Humanoid. Something doesn’t add up.
Though Musk is believed to be one of the top technology experts, with a bunch of companies under his belt, the truth is entirely different. Tesla shares have been tumbling ever since he announced his stake in Twitter. SpaceX investors in 2018 had complained that Musk is paying his employees at Boring Co. using SpaceX funds.
The Boring Co, Musk’s tunnel construction company, had been silent all this while after it started backing out from the projects it had promised in the cities. Similarly, Neuralink, his brain-computer interface, is yet to receive approval for human trials. More than just one of Musk’s problems, if X actually comes out as a superapp, there is very little reason to believe that it would actually work. Apart from WeChat in China, there is almost no success story of superapps.
Thus, it looks like Musk might have an upper hand on OpenAI and the world when it comes to generative AI capabilities because of Twitter data, the opposite is equally possible – adding another failed project to his portfolio. Musk’s AI project will either be the best, or fail miserably.