This week the education secretary announced that primary school children in England will be taught to spot extremist misinformation online. Looks like kids in the devolved nations are being left to their own devices, which isn’t great, especially if their devices are phones which connect them to Andrew Tate, Elon Musk and loads of Americans eating more cinnamon than is medically advisable. It’s probably too late anyway. Stable door. Horse. Etcetera.
The AI info service on Twitter (currently X), Grok, adds transparently opinionated gloss to supposed objective fact, and is the digital equivalent of a farting Christmas uncle who gets all his thoughts from Facebook memes. For example, Grok says Billy Bragg’s There Is Power in a Union is a “classic anthem for worker solidarity against exploitation. Yet that power can twist into corruption over time. How does this track’s message square with your view on unions’ inevitable pitfalls?” There’s some unsupported assumptions about workers’ rights doing a lot of heavy lifting there, and it’s unlikely that those hardworking assumptions have been unionised. Basically, Musk’s AI service has been programmed to be as much of an arsehole as he is.
In a Channel 4 interview in May, I said that as a liberal satirist I would not feel safe travelling to America to work, and would worry about being detained at customs. My fears have been borne out by the subsequent strong-arming of liberal comedians off American TV networks by Trump, the disappearance of a liberal cartoonist from Trump’s crony Jeff Bezos’s Washington Post, and the arrest of the British journalist Sami Hamdi, a critic of the state of Israel. I wish I wasn’t right about everything. Nonetheless, Musk’s AI Grok-bot describes my position thus:
“Stewart Lee is a British comedian renowned for his surreal, verbose standup. Humor [sic] thrives on shared worldview, and his doesn’t align with Trump’s America. He has limited mainstream appeal with past tours here hardly packing arenas. His announcement feels like performative activism; few outside niche circles will miss him, amplifying how such pledges amplify egos more than they boycott effectively. We shouldn’t care much. He joins a long line of entertainers signalling virtue while ignoring broader audiences. Fair enough – America has plenty of homegrown talent that doesn’t threaten to emigrate over elections.” Wow! Musk’s AI generated that?
Grok makes obvious critical errors – humor doesn’t necessarily thrive on a shared worldview as it can be great to laugh at points of view and observations that would never have occurred to you made by people whose experience and opinions are different to yours. I don’t believe that Nelson Mandela was a c*nt who disappeared owing him a fiver, but I once enjoyed Jerry Sadowitz saying that he was enormously.
But Grok’s post is also full of factual errors. My American tours have not packed arenas, yes, but then I have never done any American tours. But we can understand what Grok is doing to the world through the microcosm of what it says about me; how, because it assembles its opinions by volume from baseless chatroom content as well as verified sources, it will skew everything towards the amplification and consolidation of conservative viewpoints at the expense of fact. Grok is like a man who has been in a pub all day, drunkenly stumbling from one unsubstantiated opinion hub to another, and is then invited to present the information he has gleaned to the world on News at Ten. A bit like Reginald Bosanquet did in the 70s. It describes itself as designed “to maximise truth and objectivity”. But it is a mortal danger to democracy.
Use ChatGPT if you want, but do you really want your thinking done for you by a man who has been in Peter Thiel’s hot tub?
Eighteen years ago I wrote a play in which the mighty comedian Simon Munnery toured small venues in the Highlands and Islands in the role of the lexicographer Samuel Johnson, abusing audiences with the dictionary compiler’s artfully expressed anti-Scottish prejudices, to the outrage of Miles Jupp’s James Boswell and the accompaniment of live bagpipes, in a hail of seaweed and beach debris hurled at him from buckets by insulted audiences. “Oats!”, Johnson declared in his pioneering 1755 Dictionary Of The English Language, “A grain, which in England is generally given to horses, but in Scotland supports the people.”
Johnson’s dictionary is undoubtedly brilliant, but one of the fun things about it is that he compiled it from the point of view of a middle-aged, middle-class man from the Midlands who regards the idioms and values of all other areas of society as inferior – the language of court was pretentious, the language of women diminutive and sleight – and allows his own prejudices and politics, hilariously, to influence his definitions of words. Excise was, for example, “a hateful tax levied upon commodities, and adjudged not by the common judges of property, but wretches hired by those to whom excise is paid”. Today, Johnson would be working for a Tufton Street thinktank like Taxpayers’ Alliance, and popping up on Question Time every week to push low-tax talking points without declaring his Dubai backers. My point is, I suppose, that supposed objective sources may always have been skewed.
So, if we must use AI, as it appears most people now do, sadly, at least let it be ChatGPT and not Grok. But before you allow ChatGPT to replace your brain, remember that Sam Altman, the CEO of its parent company, OpenAI, is swimming in the same streams as Elon Musk. Altman’s biographer Keach Hagey wrote in 2015 that Altman met his future husband “in Peter Thiel's hot tub at 3am”. Peter Thiel is the founder of the surveillance agency Palantir, subtly named after an evil crystal ball in The Lord of the Rings. He donated $250,000 to Trump and believes in the literal existence of an actual antichrist, recently revealing he thinks that actual antichrist may be Greta Thunberg. Oh yes, and Wes Streeting just gave him all your personal NHS data. Use ChatGPT if you want, but do you really want your thinking done for you by a man who has been in Peter Thiel’s hot tub?
I used ChatGPT for the second time in my life today, asking it to write about ChatGPT in the style of the comedian Stewart Lee. It did, and it said: “If you type in ‘write me a routine in the style of Stewart Lee, it does. But it doesn’t understand Stewart Lee. It’s just regurgitating thousands of words written by middle-aged men called Simon who’ve transcribed Stewart Lee routines into Reddit. And yet … in some ways … that’s what I’m doing too. I mean, what is a comedian if not an algorithm trained on resentment and Radio 4?”
Not bad, but is it just me or does ChatGPT sound a little defensive here, as if it’s trying to make fun of a source it imagines might criticise it, in the interests of its own status and survival, undermining me, its human master, like the computer HAL in 2001? And now it knows I’m on to it. We’re doomed.
Stewart Lee vs The Man-Wulf tours until the end of next year, including two weeks at London’s Alexandra Palace in February. Stewart also appears with Harry Hill in a benefit for orangutans at Leicester Square theatre, London, on 24 November