Sabitlenmiş Tweet
I've written about race, genetic ancestry, analyses of large biobanks, and human history
#h.v8wagygagcry" target="_blank" rel="nofollow noopener">gusevlab.org/projects/hsq/#…
I'll summarize the key points here 🧵:
English
Sasha Gusev
22.6K posts
Sasha Gusev
@SashaGusevPosts
Statistical geneticist | Associate Prof at @DanaFarber / @harvardmed / @DFCIPopSci | Blogging at https://t.co/4D7UObBNdd
Katılım Ocak 2023
2.9K Takip Edilen24K Takipçiler
@anshulkundaje They probably just ran this report through the AGI on effort=medium instead of effort=ultra
English
@SashaGusevPosts Obviously the answer is AGI or is it ASI. Either way, it can do anything including solve conflicting goals at the same time on different GPUs.
English
@npparikh And then the discussion of AI innovation itself treats it like some kind of supernatural force.
English
@npparikh It's really a strange document. I don't think it was AI written but it has that AI ~feeling~ of someone stringing together pieces from other sources without really thinking about how they fit together or why they are even in there.
English
The main issue I have with the OSTP report on science is that it sets up multiple conflicting goals and provides no guidance on how they should be prioritized. For instance:
English
It's great to "Invest in Foundational Research to Drive Scientific Breakthroughs for Emerging Technologies". This is, in fact, what the NIH and NSF already try to do. A document like this could have actually set concrete priorities for what they should be doing differently.
English
In some instances these contradictions veer into the absurd, such as the goal that funding should be prioritized for research in biological sciences over life sciences. Or that universities must focus exclusively on merit but also freeze out international talent.
English
"the answer to critique and demands to “reform” the NIH is a simple 'they already do that'. ... The next response is 'if you want more of this thing they already do, than give the NIH an increase in budget. Otherwise, what specific core activities of the NIH are to be cut to pay for that?'"
Drug Monkey@drugmonkeyblog
NIH must "open scientific careers to skilled craftspeople and practitioners" drugmonkey.wordpress.com/2026/07/22/nih… via @drugmonkeyblog
English
@mattyglesias Interesting, my wife and I read it as setting up an Ashkenazi vs Sephardic contrast rather than Diaspora vs Israeli
English
Important question for non-Jews who saw the little-seen crime thriller “Tuner” — was it clear at all to a general audience that the villains are Israeli?
It was obvious to me but I feel like most people would miss this.
slowboring.com/p/the-lost-joy…
English
Sasha Gusev retweetledi
What IP, precisely, was stolen? Because copyright on LLM output is commonly assigned to the prompter?
Or should the model providers own model output? That'd make open models the *only* viable path to use coding agents?
Treasury Secretary Scott Bessent@SecScottBessent
We support open-source AI and the innovation it unlocks. But open source is not open season on American IP. When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.
English
I don't even sports but this is a fun read on analyzing running times. Interesting to think about other non-standard measurements of elite athleticism.
open.substack.com/pub/jkpritchar…
English
Sasha Gusev retweetledi
@jennbrophy7 @pkoo562 If u have deep expertise in genomic data processing with agentic platforms, DNALMs, S2F models & interested in joining us please reach out. I'll be hiring 1-2 postdocs/staff scientists for this project. Please RT 7/
English
@mathiesoniain Lol, you gotta hand it to Adams on this stuff
English
There was a big initiative in 2020 to centralize AI research within a small number of well-funded university "hubs". Now there's a big initiative to redirect funding to individual scientists. One guy digs a hole, another guy fills it up (in this case the same guy).
Andrew Curran@AndrewCurran_
In an new report expected to be released later today the Trump administration will argue that directing federal funding towards individual researchers, and research that uses AI, will accelerate scientific progress faster than funding collages and universites.
English
@SashaGusevPosts @KelseyTuoc Very few schools "build in randomness" so is your claim that all schools that don't are ignorant or malicious? Do you have children and if so do you only send them to schools that have built in RCTs?
English
I think this is the right attitude to have about Alpha School. It'd be great if they can scalably deliver these results, but many people have claimed that and failed.
John Arnold@johnarnold
I've seen many models over past 25+ years that promised to fundamentally improve K-12 outcomes. None has proven both durable and scalable. Hence my interest yet healthy skepticism about Alpha School, which has made bold claims while offering little transparency or independent validation. I strongly second this call from @rpondiscio. If Alpha works as well as claimed, open up and let ed researchers study it and policymakers learn from it. If it works even half as well as claimed, the model can revolutionize education and lead to enormous societal benefit. I hope the team at Alpha has achieved the K-12 breakthrough we've been searching for for decades. I salute MacKenzie Price and Joe Liemandt for dedicating their time and resources to creating the model. We need more experimentation in education. There is great promise in the intersection of tech and broad-scale learning. But extraordinary claims also need external validation.
English
@Jeremy241686 @KelseyTuoc Public schools should also do more randomization, but they don't have the same luxuries as a new startup pitching a disruptive technology and recruiting from scratch
English
@SashaGusevPosts @KelseyTuoc But public schools just change what they do, e.g. remove tests in favour of holistic marking or change curriculum around, all the time without careful rigor. They are not careful institutions preserving the most time tested techniques
English
@SashaGusevPosts There are basically zero businesses in America that will sacrifice 2.5million/year for the sake of advancing scientific knowledge. you can declare all businesses malicious if you want I guess.
English
Sasha Gusev@SashaGusevPosts
@KelseyTuoc @lazy_grad The dynamics for what happens in schools are fundamentally very different. In the same sense that a surgeon is forced to make non-randomized decisions on the operating table whereas a new surgery device is required to go through clinical evaluation.
QME
@SashaGusevPosts @KelseyTuoc Do you hold public schools to these standards? The public school system is *way* bigger and as a whole has *lots* of money, they could be running tons of very high powered studies if they wanted to
English
@Afinetheorem Yeah, I have no idea what "away from colleges" could even mean here. Are individual scientists going to rent their own lab space (or work outdoors)? Set up their own payroll and HR for staff? Run their own IRBs, compliance, and IT? Recruit their own patients into studies?
English
This, in my read, is a very inaccurate understanding of the OSTP report. Direct funding of researchers on long term grants (who likely are working at universities!) is a change to how we allocate grants that most innovation economists support.
Andrew Curran@AndrewCurran_
In an new report expected to be released later today the Trump administration will argue that directing federal funding towards individual researchers, and research that uses AI, will accelerate scientific progress faster than funding collages and universites.
English
@Allodoxaa @ScientificCoder They're definitely trying to spin it, as any company would, but if enterprise is going to be one of their markets this stuff is harming that and people can see through the spin.
English
