In defense of Sam Altman

It’s a classic trajectory. You start as a likable and geeky tech tyro, you morph into a squillionaire with disagreeable habits — and somewhere on the way you become loathed by large sections of the population. 

It happened to Bill Gates — remember when he was an amiable nerd making glitchy but intriguing software? Now he is a mogul apparently injecting us with nanobots. Look at the career of Elon Musk, once a move-and-break-things hero sending amazing jets into the sky. Now he’s the brooding Trumpite Satan, who has supposedly turned Twitter into a fascist hellhole. Steve Jobs of Apple nearly went through the same process, but avoided it by the devious means of dying too young.

Now the same evolution seems to be happening to OpenAI CEO Sam Altman. Except it is happening to him much faster than any of the other fallen tech titans, which is fitting, given that he works in AI, an industry where everything is turned up to eleven and “exponential” is expected.

It’s hard to remember that OpenAI has only existed for nine years. Yes, less than a decade; and it was born when Sam Altman was just thirty. As little as three years ago it was still barely known outside techie circles. Now, with the progressive unveiling of GPTs 2, 3, 4, and o, and the potentially epochal 5 on the way, it is a company worth $80 billion (at least). It has giants like Google wondering if they are doomed, and it has the rest of the world wondering if humanity itself is imperiled. 

What, then, are the charges against Altman, its CEO?  Some of them are partly justified. Elon Musk, for instance, was an early founder of OpenAI and he says the original intent was to make AI that would be freely available, to much of the world — hence the name of the company. 

Much of that has not happened. The new OpenAI guards its secrets carefully, like a dragon sitting atop its hoard. Similarly, rumors leak out of the company’s San Francisco HQ the same way they slipped out of the old Soviet Kremlin. 

OpenAI even suffers coups and countercoups — Altman was briefly deposed late last year, in a bizarre rebellion by board members, but then he was reinstalled a couple of days later after a factory floor uprising. It is difficult to know if this was because they genuinely liked their leader, or because they were worried about their stock options. Perhaps both?

Another allegation against Altman is that he is neglecting the other foundational principle of OpenAI. The idea was, in those days, that OpenAI should create “safe and ethical” artificial intelligence. The confidence that this is happening has been undermined by the recent departure of so many senior OpenAI staffers dedicated to this side of development (known as “alignment” — as we try to align the machines with human values, and make sure they don’t turn us into paperclips). The most notable departure was alignment bigwig Ilya Sutskever, who was — notably — also involved in the failed putsch.

Finally, the charge sheet against Sam Altman is one of basic ineptitude, in various ways. For instance, he is prone to making jejune statements on social media. At one point he joked on Reddit that “AGI has been developed internally” — referencing an AI in-joke. It may have been a joke, but the advent of AGI is roughly as important as intelligent aliens landing in Regent’s Park, so it’s probably not something you should joke about.

What can be said in defense of Altman?

Altman exhibited a similar cackhandedness in recent weeks with the Scarlett Johansson brouhaha. For those that need updating: OpenAI have just developed a new talking-and-seeing AI (GPT4o). It seems they wanted to actually give it ScarJo’s voice, as a hommage to her role as a seductive AI in the famously prescient 2013 movie Her. ScarJo said no, so they got another husky-voiced actress (it sounds to me like Rashida Jones, best known for TV sitcom Parks and Rec). And yet, a few days before launch, they apparently asked ScarJo again, got another No, at which point Altman decided to troll the world anyway, and announced the arrival of GPT4o with the simple word “Her.”

All of which makes Altman look stalky and obsessed (though he is gay) and definitely quite foolish: the social media post was enough to provoke ScarJo into threatened legal action. OpenAI have had to withdraw the husky voice, for now.

Given all that, what can be said in defense of Altman?

First, and obviously: he’s a human being, helping to bring the greatest technological advance in human history to fruition. He is inevitably going to do things that annoy people; he will make errors. Second, yes OpenAI have closed up — but how else do you survive, and keep your amazing tech secret? Also, OpenAI have indeed gone corporate — teaming up with Microsoft, but AI development requires billions of dollars: where else do they get the money? 

As for safety and alignment, it is difficult to distinguish between disgruntled employees, and differing philosophies — and actual safety problems. It is notable that the biggest ethical AI screw-up to date has not come from OpenAI, it happened to Google, who made an AI so “Woke” (Gemini 1.5) it was dangerously mendacious, and was instantly taken down.

Finally, consider Altman’s personal position. Sure, he is a billionaire, with extremely nice wristwatches. But is his life that much fun? AI threatens to take away millions of jobs, and these are often jobs that people love: the arts and sciences. This will make people angry, and angry people do violent things. Altman himself has spoken of the “not zero” chance of his being shot. In the last week he has admitted he can no longer eat in restaurants in his hometown, which makes him sad.

Ergo, I am inviting you to feel sorry for a billionaire. I know that is not easy, but even if you can’t pity him, then cut him some slack. He is fairly transparent: a democrat who does podcasts. Put it another way: yes, Californian tech bros can be annoying, but I’d rather this world-changing research is done in Silicon Valley, as against Suzhou, or St. Petersburg.

This article was originally published on The Spectator’s UK website.