Vinamr retweeted
If you believe that the rational thing to do is not vote, then all rational people will come to the same conclusion, and not vote. But then since all rational voters voluntarily disenfranchised themselves, the result will be affected solely by irrational voters, which is a bad outcome! Every rational person will realise this, and (by common knowledge of rationality) realise that every rational person has realised this — since they have no way to coordinate which of them will supply the necessary votes, they can't just leave it to others: otherwise everyone could reason that way, and nobody would vote. Thus rational people conclude they should vote.
Vinamr retweeted
Before the Nazis took power in 1933, the groups that leaned most strongly into Nazi or Nazi-adjacent ideology were university students (by far), professors, doctors, Protestant clergy, judges, and the nationalist press. Many teachers too, individually – not yet as a profession.
Vinamr retweeted
AGI and the EMH: what does the market tell us about AI timelines?
TLDR: Markets are not putting much probability on the development of transformative AI (aligned or unaligned!) in the next 30-50 years – as evidenced by low *real interest rates*
forum.effectivealtruism.org/…
Vinamr retweeted
"Henrich, by contrast, posits the following [time] sequence: 4. culture, 3. cooperation, 2. language, 1. intelligence. This is the exciting hypothesis that has been generating interest and attention over the past few decades."
Two accounts of human evolution, that are each other’s exact opposite. Only one can be correct.
Joseph Heath explains which one, and why:
buff.ly/35NxmmK
Vinamr retweeted
Human extinction would almost certainly be preceded by events that you can apply strict liability to. It is better to set penalties for those specific events than let regulators set policy based on something so nebulous as extinction risk.
Vinamr retweeted
...regulators the ability to block things that they (likely) do not understand or that (more likely) just happen to become the target of a mass hysteria -- like data centers. We should also not ignore the financial incentives that labs have in softening frontier competition."
Vinamr retweeted
What I wish Jenson argued: "Even if we take every doomer claim at face value, there is a solution. We can give frontier AI companies, who have the most direct knowledge of this technology, stronger incentives to care about safety (i.e. reformed liability). Instead of giving...
Vinamr retweeted
Generalization out of distribution is a central crux of the AI risk debate. But there's a double bind
If training on narrow tasks generalizes to broad capability, AIs will be powerful
But then also training on narrow alignment generalizes to broad alignment cf. @OwainEvans_UK
On the other hand, if generalization is weak, that looks bad for alignment but also constrains capabilities
Vinamr retweeted
This is one of the many ways that the professed goal of intellectual progress comes apart from the more heart-felt goal of max personal prestige.
Vinamr retweeted
I don't buy that AI is a threat to humanity for a nerdy statistical reason.
My views are informed by the statistics of things like war. If you look at deaths from human conflicts, you get a continuous spectrum from lots of small events causing a few dozen deaths to rare outliers killing tens of millions.
In my own research, I've studied complex diseases where the overall outcome reflects hundreds of genetic factors combining in different ways.
So if AI can kill billions of people, I would expect to see some version of the same thing: smaller events where one or two of the relevant causal mechanisms are operating, larger events where several combine, and catastrophic events where everything goes wrong at once.
Instead, what we seem to have is real evidence of a few suicides and isolated harms at one end, and wild speculation about complete apocalypse at the other.
That massive discontinuity makes me very skeptical that the AI catastrophic-risk story is real.
Vinamr retweeted
An alternative hypothesis:
- model performance is plateauing
- compute is getting much more expensive
- AI data centers are massively unpopular
- slowing down AI development is a way to explain slowing progress, reduce spending, and try to regain some goodwill.
- This is an effort to save the IPO not humanity.
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
Vinamr retweeted
The biggest impact AI has had on my life has been the increased frequency between "it's so over" and "we're so back" going through my head
Vinamr retweeted
Some new joiners for a total pot size of $575k!
On my side:
@robinhanson
@phl43
@SashaKaletsky
@_gabrielunger
On the fast-growth side:
@jankulveit
plus @SamuelAlbanie continuing to increase his stake 😮
benjaminmoll.com/growth_bet/
Some real diversity of opinion at Google DeepMind: @alexolegimas joins the bet while his colleague @SamuelAlbanie takes the other side increasing his existing stake!
Total size of pot now $425,000!
benjaminmoll.com/growth_bet/
Vinamr retweeted
Economists have always noted that the cost of living, and hence economic growth adjusted for inflation, is extraordinarily hard to measure, because the nature & quality of goods changes over time, vitiating a constant yardstick (how do you compare a cold cellar with a refrigerator? How much did a laptop cost in 1934?) A new method, using AI to valuate goods, shows that 20th C US economic growth was much bigger than previously estimated.
Excited to share our new working paper: how much did the US standard of living really rise over the twentieth century — and when? With @bhwittenbrink
Using 5.1 million Sears catalog listings, we build a quality-adjusted price index for consumer goods, 1900–1990. Two headline findings: growth was much larger than official statistics imply, and its peak came before WWII, not after. (1/9)
