The model trades survival probability for higher living standards.
Chad Jones, economist who'll be joining Anthropic: "In other words, with log utility it is optimal to take a 1 in 3 chance of ending human existence in exchange for a 2/3 chance of dramatically raising living standards by a factor of 55."
Anthropic has hired an economist with . . . interesting views on human survival https://ft.trib.al/EdBhNbP
Many users criticized Anthropic's hiring of economist Chad Jones over his 1-in-3 extinction risk modeling, calling the utilitarian calculations tasteless, foolish, and pretentious catastrophism.
No Digg Deeper questions have been answered for this story yet.
FT published a piece "Anthropic has hired an economist with . . . interesting views on human survival"
Stanford economist Chad Jones joined Antropic this week.
Jones’s NBER paper showed advanced AI as a bargain between faster invention and existential risk, where smarter systems may speed research but may also create losses no economy can repair.
His most jarring result uses log utility, an economics assumption that each extra gain helps less than the last, and says a 1% annual extinction risk for 40 years leaves about a 67% survival chance.
@andreamiotti This look like a claim about what you would do given particular (unrealistic) utility, it isn't that Chad Jones is endorsing making this tradeoff! (Also, the real tradeoff is about amount you delay vs risk which looks much more favorable even under log utility.)
@andreamiotti Oh well if log utility says it's optimal then I'm sold
@andreamiotti I'm not convinced it's possible to raise living standards by a factor of 55.
FT article - https://www.ft.com/content/bb04671c-4377-4231-96ef-0f8e57ed5d1b
The AI Dilemma: Growth versus Existential Risk† By Charles I. Jones*
https://web.stanford.edu/~chadj/existentialrisk.pdf
FT published a piece "Anthropic has hired an economist with . . . interesting views on human survival"
Stanford economist Chad Jones joined Antropic this week.
Jones’s NBER paper showed advanced AI as a bargain between faster invention and existential risk, where smarter systems may speed research but may also create losses no economy can repair.
His most jarring result uses log utility, an economics assumption that each extra gain helps less than the last, and says a 1% annual extinction risk for 40 years leaves about a 67% survival chance.
@andreamiotti Sounds like good odds to me
@andreamiotti Living standards for who?
@andreamiotti Knowing nothing about this person other than the quote, I assume he's making a joke about the inadequacy of the usual tools of welfare economics to price x-risk.
@andreamiotti Also Chad Jones: "an annual investment of 5% of GDP toward mitigating catastrophic risk could be justified"
That's over a trillion dollars per year
@andreamiotti 3 years into “AI” revolution and human living standards are below the Great Depression based upon all surveys and cost of living.
SLOP article of nonsense.
@andreamiotti these are models!
nowhere does he say log utility is what we should use.
these are thought experiments to reason about economic trade-offs, not real-world policy recommendations.
this is such bad faith slander.
@andreamiotti Jesus.
@andreamiotti I tell you, I'll be scared shitless if they aren't completely full of shit.
@_AashishReddy @ShakeelHashim On the 1/3 p(doom) in Jones’ paper, “we’re not sure that this is a bet that most members of the human species would be willing to take.” Huh, you don’t say!
@andreamiotti Utilitarianism is always tasteless. As practiced by fools like Jones it's also foolish.
@rohanpaul_ai Given that an AI proposes an action, can we objectively determine whether that action is admissible before it changes the world?
@299792458MaxHH @andreamiotti but he doesn’t say log utility is optimal to inform real world policymaking
@andreamiotti May I remind you that the full sentence was "Liberté, égalité, Fraternité OU la mort"?
@andreamiotti Ah, totally insufferable, the true mark of Ant.
@andreamiotti stay with me, Bentham but with hallucinations
