Elon Musk apparently plans to take on his former buddy OpenAI after destroying Twitter, and is already gathering the resources and personnel required to do so. The busy billionaire may use the resources of his several businesses to speed up the process, but there are excellent reasons to be dubious of this plan.
When Musk withdrew his support from OpenAI in 2018, which he had helped start, there have been rumours and inferences that he is developing an AI business to compete with it.
He has been open about his disapproval of their decision to prioritise profits above pure research and development as well as their effort to create ethical AI guardrails, which he appears to find to be overly “woke.”
It is unclear how this viewpoint fits into the belief that AI poses a threat to humans and should be curbed in the long run and maybe put on hold in the short term. Given how many short-lived expedient positions Musk has abandoned, that stance may already have been given up.
Igor Babuschkin, a former employee of DeepMind, and a few other developers have been hired by Musk, according to today’s Financial Times story.
The hiring of staff members and the recent transfer of Twitter’s corporate ownership under Elon Musk’s long-proposed “everything app” X Corp are signs that the ball is at least beginning to travel in the right direction.
Yet Musk could discover that running a contemporary AI firm is not easy, much like running Twitter. Due to years of trial and error, billions of dollars in investment, and the recruitment and retention of many machine learning professionals who felt underused at companies like Google and Microsoft, OpenAI is in the position it is today (which both now find themselves eating its dust, in different ways). An enormous barrier is the concentration of AI talent there and at a select few other institutes (not to mention China).
Musk’s notoriously cruel and arbitrary management style at Twitter and elsewhere may very well operate as a severe disincentive to hiring the AI specialists in the dozen different sectors he would require, even if he manages to raise the funding for this firm. Would you quit a position at maybe the most powerful tech business today, or at an FAANG-scale company that is investing heavily in the issue, to work for Musk?
To be fair, building and tuning your own huge language model is now rather simple, but that serves more as a barrier to entrance than an onramp. Why would they need a business like the one Musk is establishing if anyone can do it, as is the case with half of the IT businesses in the world? In the most general sense, a skeleton team could be able to train a GPT-4 equivalent, but there is much more to the process than that, as shown by OpenAI’s move to become more of a typical SaaS firm.
Although the value proposition of this startup AI business has not yet been revealed, if it is similar to what Musk has critiqued about others, there may be less of what he says is top-down intervention in the natural process of technologically improved free expression. If that is the case, the alleged idea to train the LLM on Twitter data (FT characteristics to persons acquainted) is an intriguing option. and even though he recently said that Tesla’s computer power would be useful for the advancement of AI, the company’s actual work in the area has constantly trailed behind promises and expectations.
There is no denying that Musk’s involvement in a sector tends to stir it up, even though this project, like many of others, may fall far short of expectations. To influence events and have a place at the table, one frequently just needs billions of dollars and not the market leader. And Musk consistently appears to have that available to him.
Subtly charming pop culture geek. Amateur analyst. Freelance tv buff. Coffee lover