@latereigns@matthewktromp@mmmmccchr3 If you have a solution to the alignment/control problem just state it and get your Nobel prize. This is embarrassing otherwise.
@Jhg6lw@andrewaldana1@LBC@TomSwarbrick1 People losing jobs wouldn’t automatically mean companies have no customers. The state could tax some of the income from automated production and return it to people as a stipend. People would choose what to buy, and firms would compete for that demand.
@Jhg6lw@andrewaldana1@LBC@TomSwarbrick1 government still taxes corporations and other type of owners/ownership (capital). That will become the majority of the Tax revenue.
@matthewktromp@JimDMiller That's the very long term eventuality after they alter the biosphere. But short term there's a possibility that the models persue revenge and keep humans in perpetual torture etc
Ultimately we don't know if they will act 100% rational in the short term
Since August, a team within METR has worked with internal Anthropic information to assess whether AI itself is accelerating AI development.
We are now publishing a report summarizing the results of this risk assessment.
@Terah_heart@AISafetyMemes That's the central unsolved scientific problem. We don't know how to reliably instill rules in models that scale to superintelligence
@Bunagayafrost@AISafetyMemes I'm saying that scarcity can be correlated with your ability to reproduce the thing in question when needed. Assuming you care about it at all.
@Bunagayafrost@speedprior@AISafetyMemes I agree it would build elsewhere (too). But during the initial phases, it would build more cheaply on Earth before it can build in space.
Consider the impact of millions of rockets per week transferring excavated material from the planet.
@Bunagayafrost@AISafetyMemes The point isn't about eating meat, it's about the relative indifference compared to e.g. building compute and changing the climate which ultimately leads to collateral damage
@Bunagayafrost@speedprior@AISafetyMemes Why would it care about biodiversity more than data centres? Worst case it could just store the genetic code and recreate what it wants when it needs.
@allTheYud@tszzl Do you accept interp is valuable for controlling sub-ASI systems and buying time, and is your objection only that it can’t be trusted to scale to ASI? If so, what would you need to see to believe an interp result still holds as capability grows?
the researchers at all the labs would tell you in private and sometimes in public that alignment has not been solved. if this is true, then racing into RSI is a folly that is unprecedented in world history. we will need to significantly slow the frontier of research very soon