Apratim Dey (@apratimdey2) 's Twitter Profile
Apratim Dey

@apratimdey2

PhD student in Statistics, Stanford University

ID: 1263220028154888193

calendar_today20-05-2020 21:28:30

61 Tweet

76 Followers

266 Following

Apratim Dey (@apratimdey2) 's Twitter Profile Photo

Indeed, every individual journey is a story with ups and downs. It takes a lot of strength to bring to public this though. Kudos!

Apratim Dey (@apratimdey2) 's Twitter Profile Photo

This is VERY concerning for me. Of course it has been like this for a long time but I was not too aware of this. We need real massive engagement to solve this issue once and for all. But not sure how that can be achieved.

François Fleuret (@francoisfleuret) 's Twitter Profile Photo

Serious take: how comes there is such a dominant silence from the humanities on what to expect from / how to shape a society with AIs everywhere.

Apratim Dey (@apratimdey2) 's Twitter Profile Photo

Very true. There isn't enough time to carefully validate anything, and too many things have been happening all at once. It's a weird cycle.

Rohan Paul (@rohanpaul_ai) 's Twitter Profile Photo

Beautiful research from Apple More thoughts stop helping once tasks cross critical depth. Thinking tokens rise, then crash, revealing compute inefficiency. So Standard LLMs beat LRMs on easy puzzles, unexpectedly. Researchers stress-test them on puzzles whose difficulty can

Beautiful research from <a href="/Apple/">Apple</a> 

More thoughts stop helping once tasks cross critical depth.

Thinking tokens rise, then crash, revealing compute inefficiency.

So Standard LLMs beat LRMs on easy puzzles, unexpectedly.

Researchers stress-test them on puzzles whose difficulty can
Apratim Dey (@apratimdey2) 's Twitter Profile Photo

This needs to be read. It is important to discuss the real value (if at all) of a bunch of technical jargon that unfortunately subjects that are derivatives of mathematics fall prey to.

Dr. Datta M.D. (AIIMS Delhi) (@drdatta_aiims) 's Twitter Profile Photo

🚨 Just published! All frontier AI models have failed “Radiology’s Last Exam” - the toughest benchmark in radiology launched today! ✅ Board-certified radiologists scored 83%, trainees 45%, but the best performing AI from frontier labs, GPT-5, managed only 30%. ❌ These results

🚨 Just published! All frontier AI models have failed “Radiology’s Last Exam” - the toughest benchmark in radiology launched today!

✅ Board-certified radiologists scored 83%, trainees 45%, but the best performing AI from frontier labs, GPT-5, managed only 30%.

❌ These results
Apratim Dey (@apratimdey2) 's Twitter Profile Photo

This is very true. A large segment of AI/ML research talks I have heard is storytelling and a bunch of demos. And people are also seemingly quite happy with this kind of research / presentation!

Thomas Bloom (@thomasfbloom) 's Twitter Profile Photo

Kevin Weil 🇺🇸 Hi, as the owner/maintainer of erdosproblems.com, this is a dramatic misrepresentation. GPT-5 found references, which solved these problems, that I personally was unaware of. The 'open' status only means I personally am unaware of a paper which solves it.

Jason Locasale (@locasalelab) 's Twitter Profile Photo

Very true. Proxies for merit have completely replaced the actual act of making scientific progress. I’ve heard that exact conversation countless times - even from many considered among the world’s top scientists. We have to work much harder to better define what scientific

Jon Barron (@jon_barron) 's Twitter Profile Photo

The current paper submission and review process seems unlikely to survive LLMs. One alternative would be to build a new process around talks: "submission" is making and giving a 30 minute live talk, and "review" is three experts watching, evaluating, and asking questions.

Lior Pachter (@lpachter) 's Twitter Profile Photo

Rattling off "Cell" "Nature" and "Science" when you talk about your work doesn't improve its quality, nor does it instill confidence in the results. Instead showcase exciting work in a transparent manner, share data and code, and provide details of the math / stats.

Anshul Kundaje (anshulkundaje@bluesky) (@anshulkundaje) 's Twitter Profile Photo

Repeat after me. FUCK citation counts. Just do meaningful work that you enjoy, do a great job making it accessible and present it honestly so folks can use it and build on it optimally. That's it.

Chaitanya K. Joshi @ICLR2025 🇸🇬 (@chaitjo) 's Twitter Profile Photo

I think the term ‘Virtual cell’ will have the same trajectory as ‘AGI’ or ‘Foundation models’ Initial opposition by rigorous scientists Bay Area + Demis are the only users --> mainstream term soon in few years (something about Overton windows)