AI/Machine Learning, Bharat and Bhartiya IT Industry

[S_Madhukar]

Some time ago I posted concerns about negative impacts of genAI. As I feared...
https://x.com/BrianRoemmele/status/1907 ... aQV9A&s=19

No wonder Elon is working to bring waifu in the physical world, everyone can buy one

On a different note, Kawasaki is working on hydrogen powered robot horse like a personal bike but might take 2050 to deliver !

https://youtu.be/Jq894NCTZJ8?si=qsDYeGK1YH9ceiYI

[Amber G.]

The world’s toughest high school math exam — the infamous 6-question, 9-hour marathon known as the IMO 2025 — took place this week. There is a post in Math dhaga.

AI tried… and stumbled.

Even the best of them, Gemini 2.5 Pro, managed just 13 out of 42 points — and that too after 32 attempts, burning $431.97 in the process. (For context: bronze medal cutoff was 19.)

So yes, folks… people like us — math folks, teachers, coaches, problem solvers — still have job security. AI may be clever, but it still can’t quite tango with Olympiad combinatorics

[A_Gupta]

@Amber G, my interest is not so much in how to solve as in how they came up with these problems, and how they tested that they are indeed suitable for the Olympiad. Any idea?

[Amber G.]

@Amber G, my interest is not so much in how to solve as in how they came up with these problems, and how they tested that they are indeed suitable for the Olympiad. Any idea?

Ah, great question! So the IMO problems aren’t just cooked up by one committee somewhere—they actually come from all over the world.

Each participating country can submit problem proposals ahead of time. These get collected into a big shortlist—usually 30–35 problems. Then, during the Olympiad itself, a jury (with representatives from each country) discusses, debates, and finally selects 6 problems to appear on the actual exam.

They look for a nice mix—algebra, geometry, number theory, combinatorics—with different levels of difficulty. The problems go through serious vetting to make sure they're original, elegant, challenging but not impossible, and that they haven’t been floating around online already.

Fun fact: some problems have even come from very young authors—there was one a few years ago where a high school student (India) submitted a problem that made it into the final 6!

This year’s (2025) selected problems, for example, came from:

USA (Problem 1 by Linus Tang)

Vietnam (Problem 2 by Quang Hung Tran)

Colombia (Problem 3 by Lorenzo Sarria)

Lithuania (Problem 4 by Paulius Aleknavičius)

Italy (Problem 5 by Massimiliano Foschi & Leonardo Franchi)

Singapore (Problem 6 by Zhao Yu Ma & David Lin Kewei)

(Yours truly has been involved in authoring or reviewing for this and similar contests—so I can say firsthand, it's a fascinating and humbling process.)

If you're curious, you can even check out the full shortlist and jury documents on the official site:
https://www.imo-official.org or similar sites for national contests for USA and India

[Vayutuvan]

@Amber G, my interest is not so much in how to solve as in how they came up with these problems, and how they tested that they are indeed suitable for the Olympiad. Any idea?

If you are looking for how to come up with problems in general, i.e. not specific to IMO, I will take a shot at it from proof theoretic point of view and constructive mathematics. It has all kinds of applications. Not just competitions.

Later ...

[A_Gupta]

Sure! In the appropriate thread

[Vayutuvan]

Sure! In the appropriate thread

Math then.