>On a composite of all three dimensions, a reviewing agent powered by GPT-5.2 scores above each paper's top-rated human reviewer (60.0% vs. 48.2%, p = 0.009)
I've agreed with this conclusion for a while: the decline in labor's share of income is mostly due to automation, broadly construed. And this has probably been a consistent trend back through human history, and beyond.
E.g. when deciding whether to cooperate with, or murder, your fellow homo erectus, murder becomes more attractive if they had a hand axe that you could just take and use yourself just as well.
The trend is just nonlinear and that is becoming more obvious as it approaches the asymptote.
I'd like to 2nd schiffern' s comment. Your comment gives a persuasive perception of competence. But that's not sufficient to accept your point. I'd love to have more data and details to work with so I can decide which viewpoint to trust here. So any sources you could provide would be really helpful. otherwise it doesn't sound responsible for me to just trust the "experts" either.
This isn't news at all, but simply a process that has gone on for decades. But it is nice to see the NYT giving it coverage. It would be wonderful if a moderate consensus could form to support reform of the process. Not gonna hold my breath though.
Also nice to see someone from Works In Progress getting visibility. WIP is the best, and you should read / listen to it:
This news seems consistent with what I've heard in general about the genetic influence on behavior (and most other traits we care about, like height): They are often very genetically influenced (on the order of 50%), but this influence is usually spread over a huge number of genes and variants. Possibly over every gene. I remember a study about height where about half the genetic influence was only captured when they considered the literal whole genome (vs. SNPs), and even then a huge amount were unique variations.
So, yes "trait X" likely has a strong genetic component. But don't hold your breath that you'll be able to control it by CRISPERing it or something. And don't hold your breath that we'll even be able to model and predict the effects of any particular genomic sequence. This is part of why I'm less worried about fertility eugenics creating a dystopia. Several other dystopia's will beat it to the punch.
Not sure what the new data here is, but I think this is an important point worth promoting. Perhaps it's the geographic association they show. In that case there could just be a confounding correlation: areas with more dynamic economies will show more mixing, and less dynamic ones will separate with the cream rising to the top (and curdling ;-P).
My guess (and experience), is the these cross-class friendships are most important for networking, and the biggest single benefit of that is that it enables people to interview for jobs that they would have otherwise been screened out of.
Since, IMHO, this is a mostly zero-sum effect (though of course there is some value in trusted references), I think it would be good to try to reduce the impact of this in our society.
> The trend reflects an end to an important chapter of the story of mass incarceration, an era in which drug laws became more punitive, the war on crime mushroomed and the number of people sentenced to time behind bars surged. States poured billions of dollars into new prisons.
> Now, the closures reflect a broader shift. Many states have revised sentencing laws and expanded education and job-training programs that help reduce the chances that people return to prison, JC Hendrickson, senior policy strategist for the Brennan Center’s Justice Program, told Straight Arrow.
> “States have similarly seen prison populations drop,” he said.
Can't get past the abstract, but that lays out the key assertion:
>Accounting for income in holding companies erases half of the difference in the 1980-2020 rise of the top 1% fiscal income share between Nordic countries and the United States, where rules penalizing the use of holding companies have been in place since the 1930s.
note this was GPT-5.2
reply