Elon Musk and the awfulness of TruthGPT

The Tesla/Twitter CEO wants to build his own AI for the MAGA faithful. Good luck with that, Space Karen.

Four versions of Elon Musk sitting on a pile of "truth." Source: Midjourney.

This is why we can't have nice things.

Three weeks after signing a letter calling for a six-month "pause" on the training of new large language models, the Chief Twit has announced plans to create his own large language model, which he has modestly titled "TruthGPT." [1]

He revealed this in a recorded interview with Tucker Carlson. His reason? Because current large language models created by firms like OpenAI are too 'woke'. Here's a snippet of that conversation [2]:

"I'm worried about the fact that [OpenAI] is being trained to be politically correct, which is simply another way of saying untruthful things... A path to AI dystopia is to train an AI to be deceptive. So yeah, I'm going to start something I call TruthGPT, a maximum truth-seeking AI that tries to understand the nature of the universe."

He also said this:

"An AI that cares about understanding the universe is unlikely to annihilate humans because we are an interesting part of the universe. [3] ...Humanity could decide to hunt down all the chimpanzees and kill 'em, but we don't.... We're actually glad they exist and we aspire to protect their habitat..." [4]

Chimpanzees watching the Tucker show would probably beg to differ. [5]

Bullshit in, Bullshit out

The luminous Nina Schick calls out Elon on this BS.

View the actual video here on Instagram.

Her point is that AI's propensity to "hallucinate" — to make up 'facts' in order to fulfill the requirements of a prompt — means it can never be expected to produce something a reasonable person would consider "truth."

But Musk is not a reasonable person, nor is he really interested in truth. You can see it in his efforts to make Twitter a safe haven for 'alternate facts,' by allowing serial fabulists and hate mongers back onto the platform, and by trying to drive legitimate news sources like the New York Times and NPR off of it.

He wants to make AI political. He wants to create a version of AI that his tribe will use instead of the ones created by the 'woke mob.' That's his target market.

Rhymes with Rump

It's impossible to ignore how similar Musk is to a certain former resident of 1600 Pennsylvania Avenue. The same desperate need for attention. The same mix of petty grievances and faux victimization, despite being born into extreme privilege. [6] The same physical vanity, despite the largely fictional hair and pudding-like physique.

Maybe a little spray tan would not be such a bad idea after all. Source: PageSix.

The same tribe of adoring deplorables. [7] The same utter perversion of the word "Truth."

Given how many AI models have been accidentally trained to produce biased results, it certainly seems possible to deliberately train AI to be biased in a specific direction. Maybe that's what Musk is hoping for -- to teach an AI model to be the opposite of 'woke', whatever that means, in the fevered imagination of MAGA-land. [8]

If TruthGPT ever comes to pass, I have serious doubts that it will produce results that are both a) usable, and b) acceptable to the population Musk is attempting to cater to.

The real truth is, we don't really know what AI models will do. But if TruthGPT or something like it fails to produce the desired results, we do know what Musk and his tribe of flying monkeys will do.

They'll figure out a way to make it someone else's fault.

Are you sick and tired of hearing things from uptight, short-sighted, narrow-minded hypocrites? All I want is the truth. Give me some of it in the comments below.

[1] I am trying to generate a sufficiently sarcastic response to that, but I'm coming up blank. Maybe I should ask TruthGPT to come up with one for me?

[2] Full disclosure: I did not watch the interview, because I am not a fan of projectile vomiting. However, I did order up a transcript and read that. It's like two prep school sophomores sitting in a dorm room smoking really bad weed at 3 am and solving all of the world's problems.

[3] He is probably stealing/misquoting a famous joke by Marvin Minsky, one of the fathers of AI: “Once the computers get control, we might never get it back. We would survive at their sufferance. If we’re lucky, they might decide to keep us as pets.”

[4] Apparently that doesn't include the monkeys that have been killed while being used to test Musk's Neuralink brain implants.

[5] This would also qualify as animal cruelty.

[6] His alleged father allegedly owned an emerald mine in Zambia. Musk has told different versions of this story while also denying it vehemently. Remind you of anyone?

[7] To paraphrase Roy Scheider in Jaws, 'We're gonna need a bigger basket.'

[8] Gabby.ai, the image generator concocted by the virulently racist Gab social network, appears to have been trained on a dataset of army and police personnel -- nearly every person and even some animals are depicted wearing military-style uniforms. It's friggin' weird, but it also demonstrates one of the problems of using limited data sets to build your AI.