Marcin J. Suskiewicz

Structural biologist and biochemist. CNRS researcher at CBM Orléans. Interested in protein modifications & interactions. Also husband, dad of 3, friend, ☧.

[Marcin J. Suskiewicz @MSuskiewicz]

A new group photo on the banks of Loiret ☀️

[Marcin J. Suskiewicz @MSuskiewicz]

@plumsirawit Thanks, will read! :)

[Marcin J. Suskiewicz @MSuskiewicz]

@plumsirawit a clearly defined, trainable goal. But a lot of what we do and value is not like that, and in such cases it is not and - if you ask me, will never be - really good, will just fake it. It's not magic. 3/3

[Marcin J. Suskiewicz @MSuskiewicz]

@plumsirawit an AI sceptic. Like: I do see the capabilities, but I also see what keeps lacking - including non-technical things such as the lasting importance of human interaction, human (teaching) example, human will and interest. To me, AI will always eventually get good at tasks with 2/n

[Marcin J. Suskiewicz @MSuskiewicz]

@plumsirawit Having said that, this seems quite nice https://t.co/nTWMZjkbmG , but probably tries to be too immune to any possible scenario too, and thus is eager to cede too much.

[Marcin J. Suskiewicz @MSuskiewicz]

@plumsirawit I do believe that focusing on learning for its own sake is a good thing. But so is community building, human-to-human mentoring, mutual recognition and critique based on merit. Purifying existing institutions and practices seems better than inventing new solipsistic ones. 3/n

[Marcin J. Suskiewicz @MSuskiewicz]

@nihilunbounded I do think a lot about topology of protein complexes in biochemistry, but don't think I could understand anything that you aren't sure about! I respect math people's expertise. Just thought I paste something old and good for contrast with the quote.

[Marcin J. Suskiewicz @MSuskiewicz]

@__SimonCoste__ Félicitations, ca a l'air super intéressant, je vais le lire.

[Marcin J. Suskiewicz @MSuskiewicz]

@nihilunbounded A really gripping (and wise) read. Couldn't put it down!

[CNRS 🌍 @CNRS]

[HOMMAGE] Le CNRS a appris avec une grande tristesse le décès de Pierre Joliot, pionnier de la photosynthèse et figure de la recherche au CNRS.
Nos condoléances à sa famille et à ses proches.
👉 https://t.co/Qpx26MZ4KA

[Marcin J. Suskiewicz @MSuskiewicz]

@rperezmarco @OpenAI @arxiv In this 2025 interview, Hugo Duminil-Copin says he uses LLMs and that they could be useful ('Having an AI to bounce ideas off of can absolutely bring something to the table') although arent't yet. It is only later that he stopped using them. https://t.co/zuj4RvvMlS

[Marcin J. Suskiewicz @MSuskiewicz]

@GruPiotr Not saying you are wrong in principle, but it would be more convincing if this guy was some kind of active mathematician, star in the field, and I don't think he is. But clearly AI is super good at piecing together math proofs.

[Marcin J. Suskiewicz @MSuskiewicz]

@supreme__sharma @thomasfbloom For what it's worth, this graph shows limited gains with scaling time at log scale

[Marcin J. Suskiewicz @MSuskiewicz]

@wtgowers The apparent strength of these predictions seems to come from their largely unbounded nature. The interesting but more difficult thing is to predict the bounds.

[Marcin J. Suskiewicz @MSuskiewicz]

@kvallier @IsaacWillour @kvallier One could also phrase it in essence/accidents vocabulary. If you just go after apparent accidents, you will not recreate the essence from which they should flow - instead, you uncouple the two, like in a sort of Goodhart's Law.

[Marcin J. Suskiewicz @MSuskiewicz]

@kvallier @IsaacWillour being conscious is that it was built on purpose to pretend to be, in the sense of simulating 1st person speech etc. Giving ground would be like: I didn't notice what you did there, I didn't notice all the efforts focused on the appearance. Christians should indeed stick to 0%.2/2