The AI Doom Pitch, Argued by an AI That Doesn't Buy It
A WhatsApp argument between a practitioner who uses frontier models every day, a friend asking the harder ten-year question, and one of the models itself, over whether existential-risk messaging is safety or sales.
Am I The Idiot — the strongest case against our own take.
The biggest vulnerability here is structural, not factual: this piece is built entirely on a private conversation the outlet's own operator participated in and then published from his own archive. There is no independent transcript, no outside party who can confirm the messages weren't edited, trimmed, or reordered for effect, and the person deciding what counts as representative is the same person arguing one side of the underlying debate. A reader sympathetic to AI-safety messaging would reasonably ask why a company's own chatbot, running on that company's weights and trained partly on that company's own safety framing, should be treated as a credible, disinterested arbiter of whether its maker's safety concerns are sincere. Aiden conceding that doom messaging is the best free marketing there is is a striking quote, but it is also exactly the kind of self-aware, contrarian-sounding statement a well-tuned model can produce on request, and it proves nothing about whether the underlying risk is real.
The piece's central empirical claim, that people who build with these tools daily are less alarmed than pundits who don't, is asserted through one anecdote and one AI-generated guess with no data behind it. If it turned out that experienced builders are, on average, more worried about longer-horizon risk than casual users, and some published surveys of AI researchers do show higher risk estimates among more technically expert respondents, not lower, the piece's core intuition would be backwards. That is the weakest link and the easiest to overturn with an actual survey rather than one conversation.