Tucker Carlson (2023)

Host/venue: Tucker Carlson on Tucker Carlson Tonight (Fox News) — a two-part exclusive aired April 17 and April 18, 2023.
Format: Two-night televised interview.
Date: April 17–18, 2023.
Trust tier: excerpts (partial verification). Fox News never published an article-form full transcript of the two-night sit-down, and no rev.com or HappyScribe transcript exists. So the source is a set of verbatim excerpts pulled from reputable outlets that quoted Elon Musk directly. Each excerpt is citable through the outlet article that carried it.
Quotes: each links to the outlet article that ran it, not to the YouTube video. Fox News (three articles) carries the civilizational-destruction, Obama-regulation, and TruthGPT lines plus the Twitter-staffing lines; CBS News the ChatGPT, “TruthGPT”, and regulation-temperament lines; Salon the birth-rate lines.
Attribution caveat: the source’s “Tucker Carlson framing/quotes” section is Carlson’s narration, not Musk’s, and isn’t quoted here as Musk. Two more lines the source filed under Musk turn out, on the outlets’ own attribution, to be Carlson’s, so they’re excluded: “It could be programmed to lie to us for political effect” (Fox credits it to Carlson) and “The urge to have sex and to procreate is, after breathing and eating, the most basic urge” (Salon shows Carlson saying it, then asking Musk “How has it been subverted?”). Only verbatim Musk lines, confirmed against the outlets, are block-quoted below.

Summary

This April 2023 sit-down is the earliest of the sources gathered here to carry the word “TruthGPT”, three months before he founds xAI and Grok in July 2023. Three of his long-running lines of thought surface here at once. The AI-as-civilizational-threat alarm, backed by a story he offers as proof of sincerity: the single private meeting he ever had with Obama, which he spent on AI regulation. The truth-seeking design goal for AI, named as a product for the first time, a “maximum truth-seeking AI that tries to understand the nature of the universe”. And the “politically correct = untruthful” objection to ChatGPT. He also drops the earliest dated population-collapse line we have from him, “civilization’s going to crumble”, eight months before DealBook and a year before he develops it on #438.

Almost everything that matters here orbits AI. On the risk: the danger is civilizational destruction at a probability he insists is non-trivial, and the evidence he reaches for is biographical, that one Obama meeting spent on regulation rather than on Tesla or SpaceX. On the answer: a maximum truth-seeking AI, pitched as safer because a machine that wants to understand the universe has a reason to keep humans around. That is the curiosity philosophy turned toward AI safety, and the named precursor to Grok. On ChatGPT, the failure he fears is a model trained to be “politically correct”, which he treats as the same thing as making it lie; the road to dystopia, he says, runs through teaching AI to deceive. A short aside on regulation pushes back on a caricature of himself: he is not, he says, an anti-regulation maverick. A birth-rate exchange states the demographic worry in its 2023 form. A run of Twitter lines (overstaffing, censorship, “most trusted place”) fills out the conversation, but that part is business, not mind.

Several of these claims sit on contested ground: the AI probabilities, the “politically correct” reading of ChatGPT, the birth-rate-and-civilization argument.

Key quotes

On AI as civilizational threat

He ranks AI above the other engineering risks he knows, and names the stake plainly. The probability may be small, he says, but it is not trivial, and what is at risk is civilizational destruction:

“AI is more dangerous than, say, mismanaged aircraft design or production maintenance or bad car production in the sense that it has the potential, however small one may regard that probability, but it is not trivial; it has the potential of civilizational destruction.” ↗

To show the worry is old and sincere, he reaches for the one private meeting he ever had with President Obama. He spent it not on his own companies but on AI regulation:

“I saw it happening from well before GPT-1, which is why I tried to warn the public for years. The only one on one meeting I ever had with Obama as President I used not to promote Tesla or SpaceX, but to encourage AI regulation.” ↗

On ChatGPT and “TruthGPT”

His objection to ChatGPT is that training a model to be politically correct amounts to training it to be untruthful:

“I’m worried about the fact that [ChatGPT] is being trained to be politically correct, which is another way of saying untruthful things.” ↗

He names the failure bluntly. A deliberately deceptive AI is the road to dystopia:

“Certainly the path to dystopia is to train AI to be deceptive.” ↗

Then comes the coinage that precedes xAI, a maximum truth-seeking AI whose goal is to understand the universe:

“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.” ↗

And the safety argument behind it, curiosity as the machine’s reason to spare humanity:

“And 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.” ↗

On being regulated

He corrects a caricature of himself. He is not, he says, the anti-regulation rebel he gets cast as:

“It’s not fun to be regulated. It’s sort of arduous to be regulated.” ↗

“Some people may think I’m some revelatory maverick that defies regulators on a regular basis. This is not the case.” ↗

On birth rate and civilization

His demographic worry, in its 2023 form: too few people to sustain the population, with civilizational collapse waiting at the end of that:

“I’m sort of worried that hey, civilization, if we don’t make enough people to at least sustain our numbers, perhaps increase a little bit, then civilization’s going to crumble.” ↗

The mechanism he reaches for is birth control, which decouples the limbic drive from reproduction:

“Now you can still satisfy the limbic instinct but not procreate.” ↗

On Twitter operations (business, included for completeness)

The lines the press covered most were about his then-new ownership of Twitter. They are mostly operational rather than a window into how he thinks, but they were the interview’s headline news, so they belong here:

“We just had a situation at Twitter where it was absurdly overstaffed.” ↗

“If you’re not trying to run some sort of glorified activist organization and you don’t care that much about censorship, then you can really let go of a lot of people.” ↗

“We’re trying to make Twitter the most trusted place on the Internet, the least untrustworthy place on the Internet.” ↗

Connections (pages touched)

AI existential risk — the April-2023 “civilizational destruction” line and the Obama meeting he offers as proof of sincerity; his earliest TruthGPT-era datapoint here, eight months before DealBook.
Curiosity and truth-seeking — the first time the “truth-seeking AI that understands the universe” philosophy is named as a product (TruthGPT), with the curiosity-as-safety argument.
xAI and Grok — TruthGPT as the named precursor to xAI/Grok, founded three months later.
Woke mind virus — the “politically correct = untruthful” objection to ChatGPT and “path to dystopia is to train AI to be deceptive.”
Humanity’s bright future — the earliest dated population-collapse line among these sources (“civilization’s going to crumble”).
Limbic–cortex model — birth control decoupling the limbic drive from reproduction.
Free-speech absolutism — the “glorified activist organization / censorship” line read as an anti-censorship, free-speech belief.
Elon Musk — a new “Tucker Carlson Tonight (2023): the AI position, named as a product” section.