Sayash Kapoor
@sayashk
Followers
10K
Following
4K
Media
159
Statuses
1K
CS PhD candidate @PrincetonCITP and senior fellow at @Mozilla. I tweet about agents, evaluation, reproducibility, AI for science. Book: https://t.co/tb2lXSP2gB
Princeton
Joined March 2015
The mainstream view of AI for science says AI will rapidly accelerate science, and that we're on track to cure cancer, double the human lifespan, colonize space, and achieve a century of progress in the next decade. In a new AI Snake Oil essay, @random_walker and I argue that
13
61
228
RT @evijitghosh: New blog post alert! 🚨"What is the Hugging Face Community Building?", with @YJernite and @IreneSolaiman . The AI narrative….
0
4
0
RT @random_walker: If we compared AI capabilities against humans with no access to tools, such as the internet, we would probably find that….
0
27
0
RT @RishiBommasani: In the running for my favorite blog post from Sayash and Arvind!. When people ask me for areas I am most excited about….
0
4
0
@random_walker Per-capita research slowing down is one thing, but by many metrics even aggregate research is slowing down or constant. (We summarize these in this table.) . We don’t think this is inevitable, and there are many interventions worth considering.
@sayashk @random_walker I wholeheartedly agree with the sentiment of this post! However, if the number of papers has increased 500 fold, and the average disruption of a paper has decreased 10 fold, doesn't that still suggest that 50 times more disruptive discoveries have been made?.
0
3
12
RT @random_walker: We ourselves are enthusiastic users of AI in our scientific workflows. On a day-to-day basis, it all feels very exciting….
0
12
0
RT @random_walker: Some aspects of AI discourse seem to come from a different planet, oblivious to basic realities on Earth. AI for science….
0
74
0
RT @kennylpeng: Are LLMs correlated when they make mistakes? In our new ICML paper, we answer this question using responses of >350 LLMs. W….
0
46
0
After we invented the dynamo, it took us 40 years to electrify factories. In the process, we had to redesign the entire factory layout — electrifying existing factories didn't cut it. Software engineering will likewise need to undergo drastic changes to truly benefit from AI.
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
14
33
194
RT @METR_Evals: We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The resu….
0
1K
0
RT @snewmanpv: How much time do AI coding tools save? @METR_Evals just released a rigorous study with a startling result: developers take 1….
0
11
0
RT @CarnegieIndia: 🎙️ New #InterpretingIndia episode!. @NidhiSinghLive joins @sayashk to explore the hype, hope, and hazards of artificial….
0
2
0
RT @daniel_d_kang: As AI agents near real-world use, how do we know what they can actually do? Reliable benchmarks are critical but agentic….
0
29
0
RT @jordanmcgillis: AI tools can detect truck driver fatigue and prevent deadly crashes. But the Teamsters are blocking their rollout. M….
0
79
0
RT @random_walker: When coding with agents, my ideal GUI for context engineering would look like this. Key features:.* Visually pick, resiz….
0
9
0
RT @random_walker: The origin story of “AI as Normal Technology”, and lessons learned. Many people have asked how the “AI as Normal Technol….
0
21
0
RT @random_walker: A post by Stripe engineer @thegautam on building a successful payments foundation model for fraud detection recently wen….
0
18
0