What could go wrong with Elon Musk's AI? Well, everyone could die

Facebook, Google, Salesforce, Twitter, Cisco, Amazon, Intel, Tesla – only Apple has managed to escape the vicious jobs cull of the post-pandemic “tech wreck” that has swept across Silicon Valley.

It is estimated that nearly a quarter of a million people employed by the big US tech firms have lost their jobs in the last year alone. Big tech stocks collectively lost nearly $4 trillion (£3 trillion) in market value in 2022. Venture capitalists who piled into the sector when the valuations of private start-ups were near their peak will suffer a silent crash.

Amid the fallout, us mere mortals might expect a back-to-basics mentality or period of soul-searching – some pause for thought, at least, while the dust settles.

Not a bit of it. The technology industry is a turbocharged, ego-driven world dominated by a handful of mega-billionaires with messiah-complexes and short attention spans. That means a relentless focus on the next big thing.

Having tired of self-driving cars, flying cars, and electric cars; conquered space – or in Mark Zuckerberg’s case, gone totally Awol in the metaverse; and all-but given up on virtual reality, the tycoons of "big tech" are suddenly obsessing about the potential of artificial intelligence (AI), a development that should have us all building bunkers in the back garden and preparing for the impending apocalypse.

The latest to leap aboard the bandwagon while it still has wheels is none other than Elon Musk, a man who you would’ve thought had his hands sufficiently full trying to prevent Twitter from becoming the most costly ego-trip of all-time.

Tesla’s controlling shareholder has also started a massive price war in China’s car market, in a move that seems destined to drive some of his smaller rivals quickly out of business.

Still, that hasn’t stopped Musk from hatching plans to join the rapidly-escalating AI arms race alongside companies like OpenAI, Google and Microsoft. Yet, in true Musk form, his ambitions are muddled to say the least.

In an apparent attempt to quell concerns about AI’s darker side, Musk told Fox News: ”I'm going to start something which I call TruthGPT, or a maximum truth-seeking AI that tries to understand the nature of the universe".

“I think this might be the best path to safety in the sense that an AI that cares about understanding the universe is unlikely to annihilate humans because we are an interesting part of the universe,” Musk said.

It is an admirable quest, yet simultaneously worrying on several levels. Firstly, it’s impossible to overlook the irony of someone who has at times struggled with what’s real and what isn’t, suddenly wanting to become the bastion of the truth.

After all, this is a man who thought nothing of sharing a false conspiracy theory with his 110m followers about the life-threatening attack on Nancy Pelosi’s husband peddled by a Right-wing website with a history of spreading fake news.

Meanwhile, Musk’s attempts to portray himself as a free speech crusader have been badly undermined by the proliferation of disinformation on Twitter since his $44bn takeover last year.

The entrepreneur can’t even make up his mind on whether he thinks AI should go ahead or not. It was only last month that he added his name to a list of around 1,000 signatories from the business, academic and tech worlds, calling for a pause on the building of AI systems because the technology is “getting out of control”. Yet here he is promising to spearhead the movement.

Musk is too impulsive, too reckless, and too ideologically driven to be trusted with something like this. In his hands, the “truth” risks quickly becoming the truth according to Elon Musk.

But assuming he is serious about becoming a major AI player, the most terrifying aspect of the Fox News interview is that the best Musk could offer in terms of reassurance is that his version “hopefully does more good than harm”.

I’m sorry, but what? “Hope” is not a word that should ever appear in the same sentence as artificial intelligence. It’s all very well going on national television to talk vaguely about wanting to create something that will understand the universe but not if in the same breath you essentially admit to not understanding how it works or whether it can be controlled.

In its most benign form, AI has the potential to trigger exciting and positive change. Yet the problem is that even the most brilliant scientists on the planet can’t guarantee that it won’t mushroom from its current chatbot form into a terrifying force so destructive that it threatens mankind.

Google boss Sundar Pichai has described AI as “the most profound technology humanity is working on – more profound than fire or electricity or anything that we've done in the past.” Yet, he has also called it a “black box” and admitted that rapid advances in AI keep him “up at night”.

The most frightening prophecy has come from Eliezer Yudkowsky, a decision theorist and the lead researcher at the Machine Intelligence Research Institute in Berkeley, California.

Writing in Time magazine, Yudkowsky warns that “the key issue is not ‘human-competitive’ intelligence; it’s what happens after AI gets to smarter-than-human intelligence”.

He says that many researchers including himself expect that the most likely result of building “superhumanly smart AI” is that “literally everyone on Earth will die”.

Yet Silicon Valley is ploughing on regardless, and without a regulatory body to police the technology, no one else it seems is willing to actually step in.

Yudkowsky’s solution? “Shut it all down” before it’s too late.