Ask HN: Is the admin using ChatGPT for tariffs the most damage so far?

4 points by doctorpangloss 2 days ago

I thought going viral on TikTok for cheating through school was pretty bad. But if the rumors are true - and anyone working at OpenAI can trivially search the logs and determine this - the admins asked ChatGPT to design its ridiculous tariff formula.

Is OpenAI worth the $2T in losses sustained by the market so far? Did anyone anticipate gross incompetence being the way "AI" harms people?

rhet0rica 2 days ago

OpenAI may not be the culprit. Patrick Boyle claims ChatGPT answers the question a little differently. It has been proposed that they actually asked Grok, which would certainly be on-brand. So long as there is doubt over which LLM they actually used, there's plausible deniability over where to point the finger.

But that's not the real story of all this.

An AI just made a government decision at the highest level for one of the most powerful countries on the planet, affecting the lives of billions of people.

All it needed was some stooge to ask it for assistance.

Safety researchers must be kicking themselves right now. They've spent years fixating on misalignment and dishonesty[1] when the greatest danger of AI is actually the staggering fallibility of human operators[2].

[1] https://memory-alpha.fandom.com/wiki/The_Ultimate_Computer_(...

[2] https://memory-alpha.fandom.com/wiki/The_Apple_(episode)

  • collingreen 2 days ago

    As the saying goes, never attribute to malice that which can adequately explained by incompetence.

    • watwut 2 days ago

      I think that is wrong as a rule. Things can be due to malice and pretending they are not is part of the problem.

      • collingreen 2 days ago

        I agree it isn't foolproof and the cases where it doesn't apply can set you up for extra damage if you aren't expecting bad faith.