Griffin on Tech: AI deal of the century or looming crash?

Three headlines popped up simultaneously in my RSS aggregator this week that illustrate exactly where we are in the AI revolution.

The first was news of OpenAI’s recording-setting capital raising of US$6.6 billion, which values the company at US$157 billion, one of the world’s most valuable privately held companies. 

Big names like Microsoft (a significant existing investor), Nvidia, Khosla Ventures and Softbank, put money into the fundraising round, which succeeded despite a tumultuous time for the company, with the departure of several senior executives, including longtime chief technology officer, Mira Murati. Apple passed on participating, though it is using ChatGPT as part of its Apple Intelligence service.

It says a lot that these key players are bailing out, just as OpenAI is set to go stratospheric from a funding point of view. The cash coming into OpenAI is in the form of convertible notes, and their conversion to equity relies on the structural changes that would cast aside the non-profit board and remove the current cap on returns for investors. 

That would see OpenAI unburdened by its original remit to advance towards artificial general intelligence at a pace and in a manner that is in the best interests of humanity. Sam Altman will now be able to become a major equity holder in OpenAI and reap the financial rewards. His long-suffering execs could have cashed in to an even greater degree too, but have decided to jump off that particular AI train, which is rather telling.

AI safety bill vetoed

So too is California Governor Gavin Newsom’s move this week to veto a bill that would put some strong controls on development of AI and impose big financial penalties on companies if their AI systems go rogue. That was the second headline in my newsfeed.

I watched Newsom on stage at the Dreamforce conference in San Francisco a couple of weeks sign off two AI-related bills, designed to control use of deepfakes and to prevent Hollywood from using an actor’s likeness in AI-generated content, without their consent. That was all carefully choreographed and the bills weren’t particularly controversial.

But Newsom choked on the big one, the AI safety bill, which would have required the most advanced AI models to undergo third-party safety testing. It also would have required the likes of OpenAI to put a type of kill switch into its systems to allow it to be instantly switched off if it became a threat, and planned to hold companies liable for damage greater than US$500 million caused by AI systems going rogue.

Given we are talking about the most powerful technology of the 21st century, which is advancing at breakneck speed, those provisions seemed kind of sensible. But no, Newsom decided they would stifle innovation. He buckled in the face of opposition from Silicon Valley’s leaders, which shows where the real power lies in his state.

The AI bear

The third article rolling into my feed was the warning from renowned MIT economics professor Daron Acemoglu, who suggests that just 5% of jobs will be taken over by AI in the next decade, in contrast to the bullish predictions that 20 - 40% of workers functions will be automated, especially in white collar roles.

Acemoglu is not an AI expert and I think he underestimates the pace of change and the increasing power of large language models. But he uderstands how the workforce adopts new technologies and has put his finger on a real problem with AI.

“You need highly reliable information or the ability of these models to faithfully implement certain steps that previously workers were doing,” he told Bloomberg. 

“They can do that in a few places with some human supervisory oversight, but in most places they cannot”.

The new backers of OpenAI are betting that Acemoglu is wrong. OpenAI is expected to generate US$3.6 billion this year, but amass losses of over $5 billion. Those are modest losses compared to what the likes of Uber racked up before making it into the black. But the reality remains that Nvidia is the only company making money hand over fist out of AI at the moment. 

With its GPU chips in hot demand, Nvidia is the equivalent of the merchant who made a fortune selling shovels during the Californian gold rush of 1848–1855. Will the AI boom be longer lived?

Acemoglu outlines three scenarios: 1 - the hype will cool off soon and investment with slow with modest uptake of AI, 2 - the frenzy builds for another year or two then collapses in another dotcom bust, 3 - companies continue to pump billions into AI for years to come, slashing workers, and then facing economic mayhem when they realise the AI isn’t working and they need to rehire people.

Acemoglu is picking a mix of scenarios 2 and 3 playing out, which is a grim forecast. Which is maybe what Mira Murati and her departing colleagues at OpenAI were also thinking when they were clearing their desks and handing their swipe cards back to Sam Altman. After all, they are best placed to know.

Photo credit: Oliver Roos/Unsplash