Dietmar Fischer
👤 SpeakerAppearances Over Time
Podcast Appearances
Yudkowsky forces us to ask the less fashionable question, control.
Do we understand what the system is doing?
Can we predict its mistakes?
Can we stop it?
Do we know what data it can access?
Do we know who is responsible when it acts?
That is the practical version of the alignment problem.
In daily business use, alignment means making sure the AI serves your actual goal, not just the prompt.
At the civilization level, alignment means making sure powerful AI systems remain compatible with human survival.
Same problem, very different size of headache.
Yudkowsky is also interesting because he comes from the rationalist world.
He helped shape Less Wrong, a community focused on clearer thinking, probability, bias and decision-making.
He also wrote Harry Potter and the Methods of Rationality, where Harry Potter approaches magic like a scientist rather than a lucky child with excellent branding.
That matters because Yudkowsky's AI warning is not just technical.
It is also about human overconfidence.
Humans are very good at saying, it will probably be fine.
This is not a safety plan.
It is a mood.
His value is that he attacks that mood.
He asks, what if we are wrong?