Silicon Valley’s techies are building the AI that could replace them

A frisson of fear and intrigue rippled through the Chase Center, an 18,000-seat sports and entertainment arena on the San Francisco waterfront.
Mark Zuckerberg, the world’s third-richest man, sat on stage in a loose-fitting shirt and sporting a mop of curls, to hold forth about his 20 years at the helm of Meta, the owner of Facebook, WhatsApp and Instagram. Legions of nerds, start-up founders, investors and Silicon Valley worker bees hung on his every word. It was when the conversation — a live recording last week of the popular Acquired technology podcast — turned to how he views the company and its employees that those ears started ringing.
Zuckerberg extolled the brutal downsizing he led in 2022 and 2023 at Meta, when he got rid of 21,000 workers in short order. “You wanna keep things lean,” he said, pining for the days when “you could just have meetings in your own head about what direction you want to go in”.
Eighteen months on from that epic bloodletting, Zuckerberg made it clear he was still very focused on transforming the social media behemoth into “the leanest version of a large company that we can be”. He added: “It is problematic when you have more people working on something than you should.”
You could almost hear a collective gulp from the rank-and-file technology workers in the crowd. Their industry is coming out of a brutal period, and not just at Meta. After their halcyon days during Covid, more than 470,000 tech workers have been let go, according to data compiled since 2022 by the website Layoffs.fyi. Meanwhile, venture capital funding for start-ups — in sectors other than artificial intelligence — plunged to $167 billion (£126 billion) in the first six months of this year, half the sum invested during the same period in 2021.
And yet last week, Zuckerberg was still talking about tweaking the “dial” to find that sweet spot between cutting too much and being just that tad overfed. His words carry immense weight. For many start-up founders and investors, he is the platonic ideal of a start-up founder: a technically brilliant engineer who is also manically focused on winning, a founder who turned his crazy teenage idea into a $1.3 trillion empire, where he still sits on the throne.
He delivered his “love the lean” message amid the heart-palpitating rise of artificial intelligence, which most of the people in that arena will have agreed is the most titanic tech shift since the internet. More than 50 per cent of the 1,800 AI start-ups that last year received at least $1.5 million each in funding were based in San Francisco and Silicon Valley. Many of them are promising to revolutionise everything from call centres to law, to what it means to be human.
San Francisco start-up HeyGen creates high-fidelity video clones of anyone who uploads as little as one minute of video footage. And on the street, driverless robotaxis ferry tens of thousands of people every week.
In pockets, the future has already arrived. The result is a simmering paranoia that is particularly acute on the West Coast because, of course, this is where the tools that threaten to supplant many types of human work are being developed. If one were to be cruel, it could said that tech workers are the authors of their own misfortune.
All of which leads to perhaps the most important question: do any of these AI tools work well enough to actually replace people, or to so dramatically augment human capabilities that firms need far fewer people to produce the same output? It is a difficult question to answer. David Cahn, an investor at venture capital firm Sequoia Capital, a backer of tech enterprises from Apple to Zoom, has publicly warned of the inflating AI bubble. He claimed that the hundreds of billions of dollars being pumped into AI chips and data centres by the Big Tech companies meant that they would have to produce at least $600 billion in AI-derived profits to justify the outlay.
“Outside of [OpenAI’s] ChatGPT, how many AI products are consumers really using today?” he asked. “Consider how much value you get from Netflix for $15.49 a month or Spotify for $11.99. Long term, AI companies will need to deliver significant value for consumers to continue opening their wallets.”
His criticism bubbled back up to the surface last week when it emerged that OpenAI was in talks to raise $7 billion from a consortium led by MGX, a United Arab Emirates-backed investment fund, at a $150 billion valuation. It was an extraordinary figure for a non-profit company that only five years ago created a for-profit arm. But it also emerged that the company has surpassed 11 million monthly subscribers. Combined with the licensing of its AI models to third parties, OpenAI is on track to bring in $4 billion in annual sales.
Yet it is also burning through the billions nearly as fast as it can raise them — that $7 billion would bring its total fundraising to more than $20 billion — due to the computing costs incurred to run its tools.
Meanwhile, anecdotal evidence is growing of big companies dialling down their spending on AI tools that are proving too bug-ridden, prone to giving bad information, and too unreliable to replace legacy systems.
Silicon Valley’s default setting, however, is near-delusional optimism, so investors focus instead on the cases where it is already moving the needle. Sebastian Siemiatkowski, founder of the “buy now, pay later” giant Klarna, said last month that he expected to halve the company’s 3,800-strong workforce by using AI to automate roles. And this came months after he had boasted of a single chatbot — one built by Klarna atop ChatGPT — that did the work of 700 people.
Last week, Bank of America released the results of a survey of 150 of its equity and market strategists who cover a universe of more than 3,000 companies. The analysts were evenly split on whether AI will be a net creator or eliminator of jobs. But they all agreed that the pace of innovation was quickening, and that the payoff would come far sooner than the delayed revolution ultimately unleashed by the internet. The bank said: “Chatbots that were as helpful as an 18-year-old intern one year ago are now displaying intelligence closer to a college graduate’s, proving that AI investment is producing useful apps.”
Central to that revolution is Zuckerberg, who appears to have moved to a new phase of his life. He wears his hair long. He is partial to long gold chains. He recently took delivery of a 287ft superyacht, called Launchpad. Zuckerberg is also far less interested in apologising for his company’s role in societal upheaval, and instead he is powering full steam ahead — and when it comes to AI, he plans to be right on the front line. Meta has released the latest iteration of its Llama AI model, which is more powerful than recent versions of ChatGPT, though not its newest version, and, crucially, is free for others to use — or for companies to build new bots that get relentlessly better at doing stuff that were once the preserve of human activity.
Life could get lean for us all.