Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)R

ranzispa

@ ranzispa @mander.xyz

Posts
4
Comments
376
Joined
10 mo. ago

  • Protein folding is not solved yet

    Jump

  • Agreed, I still do hope they can maintain some of their promises. However until now, I have not really seen any real advances towards making something useful.

    I do not have a deep knowledge of quantum computers, but I know plenty people working on them and often get to talk about it.

    I know people working on chemical problems who are basically approximating atoms to point charges. And either way those calculations are slower than on a CPU. For the uninformed, in chemistry the interactions between electronic orbitals is fundamental; this is in no way an approximation useful to obtain any kind of information.

    This is fine, I understand methodologies take time to develop; however as far as I understand it those techniques they're using are mathematically limited to using point charges: no matter how much they improve them that'll be the highest level of accuracy.

    I hope someone finds a way to handle such things better: as much as you can make a great machine learning model you're always depending on available data.

  • Protein folding is not solved yet

    Jump

  • Yes, X-ray is the gold standard. Technology has advanced in the sense that the protein crystallization is now more standardized and automated, as well as the analysis of the results.

    It is not the only technique, for example there are cheaper ones based on mass spectrometry which do not resolve the full structure but allow to understand which amino acids are spatially near; such information is useful when developing a protein model and to validate whether a model is plausible.

    The other two major techniques for structure resolution are NMR spectra analysis and the fairly novel technique of cryo electro microscopy.

    These in general do not resolve the protein structure to the same resolution as X-ray but have other advantages: they allow you to observe the protein structure when in solution, which may be significantly different from the crystallized structure.

  • Protein folding is not solved yet

    Jump

  • I doubt there is a comparable correctness metric between LLMs and protein structure prediction models.

    You can measure how many times they correctly predict a thing, but results will greatly change according to what your objective is. Those are only comparable when you're trying to predict the same thing.

    As such my reply would be: sometimes more incorrect sometimes more correct. However, in general, a mishandled incorrect protein structure prediction is way more expensive than an LLM hallucination.

  • Protein folding is not solved yet

    Jump

  • The interesting part would be to a reliable and computationally accessible way to handle disordered proteins. Seeing how they can move could be quite revolutionary.

    I had to work on some disordered proteins and you're pretty much just guessing, plausibly you're better off going to a casino blindfolded and play blackjack.

  • Protein folding is not solved yet

    Jump

  • AlphaFold is incorrect in plenty occasions.

  • Protein folding is not solved yet

    Jump

  • Not solved, no. Definitely much much better than before. The difference AlphaFold made is significant: we're talking about getting a decent model in a couple minutes using a PC compared to several months of calculations before.

    However we still need experimental data: in many occasions AlphaFold gives an incorrect model. With some experimental data that model can be improved, but we still have no reliable way to know what the structure of a protein is starting from the amino acidic sequence without extensive experimentation.

    That's the big promise of quantum computers, there are however two major problems in my opinion:

    1. There's still no theoretical framework which explains how once we have a quantum computer we may tackle protein folding
    2. Plenty quantum computing companies closed shortly after AlphaFold was published since they lost all funding because protein folding was "solved"
  • Science Memes @mander.xyz
    ranzispa @mander.xyz

    Protein folding is not solved yet

  • Nobody Asked for AI

    Jump

  • What I'm saying is that also LLMs grew organically. Machine learning developed organically through the years and LLMs are one of the products of such development.

    Machine learning algorithms were absolutely desired and wanted, hundreds of thousands of people worked on developing them. It is a very useful technology. If I may list one very useful development: AlphaFold, for which the 2024 Nobel prize was adjudicated.

    Now, whether LLMs are useful in chemistry and biology: kind of. Coming up with the design of a bacteriophage is not the first use which would come to my mind. Of course LLMs are being tested for their applications in such domains, but results are in general worse than dedicated models. One area in which I see value is the use of agents in orchestrating tools and analyzing results. Not really in producing new things on their own.

  • Nobody Asked for AI

    Jump

  • I come from a chemistry background. Chemistry is one of the first field in which machine learning was first applied. In fact many of the first algorithms and models have been developed by chemists.

    This has been a gradual adoption throughout the past 50 years.

    How is this different from your description of the internet history? If anything, internet was much faster than that: it took less than 50 years from the first computer to the invention of internet.

    And I'm not even considering the first developments in machine learning, which happened pretty much as soon as computing machines were invented.

  • Ecosystem services

    Jump

  • ~100x it's a logarithmic scale

  • Nobody Asked for AI

    Jump

  • I did not read the full article since it is paywalled.

    But well, nobody asked for the internet either.

  • Nah it wasn't him. He was with me studying about Jesus.

    Jump

  • That is probably one of the things the Bible would talk about if Jesus was born in our times.

  • Nah it wasn't him. He was with me studying about Jesus.

    Jump

  • Is corruption not common in your country?

  • Mathematicians are grappling with the possibility that AI might eclipse them

    Jump

  • Indeed peer reviewed exists in mathematics. Now, whether it is common practice to publish stuff on arxiv and leave it there is another thing.

    I don't know about mathematics, but I know plenty other Fields where it is common practice to just publish on arxiv.

  • Mathematicians are grappling with the possibility that AI might eclipse them

    Jump

  • Two researchers came to the same solution to a problem. In my books that's better than peer review.

  • Mathematicians are grappling with the possibility that AI might eclipse them

    Jump

  • Mathematicians are grappling with the possibility that AI might eclipse them

    Jump

  • Scientists don't often publish when they confirm an article is correct. Knowing a few mathematicians, probably they see no need to do that. They checked the proof, it was ok and that's it.

    Either way, many of those proofs come with a computer program which checks and confirms the proof is correct.

    I trust that an expert mathematician talking about such things has reviewed a few of those articles and has checked the proof.

    You may not do that; check the proof yourself or pay a mathematician to do it for you.

  • Mathematicians are grappling with the possibility that AI might eclipse them

    Jump

  • Yes, scientific articles are expensive. I know that, that sucks. That's why most of this stuff is on arxiv.

    I linked to evidence that mathematicians are using LLMs to find proofs and publish those proofs. Which is what I said is happening.

    https://academia.stackexchange.com/questions/221183/can-i-publish-a-novel-theorem-which-was-proven-with-ai-assistance

    I am not a mathematician, thus I don't really know where to find indipendent confirmation or even how that is generally handled by mathematicians. However: there are plenty proofs on arxiv that disclose have been found with LLMs. Some of these proofs relate to famous problems and have been in the news. Fields medal winners discuss the importance of LLMs and how it may produce too many proofs for humans to handle.

    I trust that those proofs published on arxiv have been reviewed by many mathematicians, if they were incorrect that would have rapidly become known.

  • Mathematicians are grappling with the possibility that AI might eclipse them

    Jump

  • Mathematicians are grappling with the possibility that AI might eclipse them

    Jump

  • Articles with proofs have been published, both by AI companies as well as by mathematicians who disclosed the whole development of the proof was done autonomously by LLMs.

  • Today I Learned @lemmy.world
    ranzispa @mander.xyz

    Just found out tracking pixels have a 3-5x 60 days ROI

    improvado.io /blog/what-is-tracking-pixel
  • Science Memes @mander.xyz
    ranzispa @mander.xyz

    Do your part in science: Be Reviewer 2

    doi.org /10.1111/medu.15679
  • linuxmemes @lemmy.world
    ranzispa @mander.xyz

    False - do nothing unsuccessfully

    man7.org /linux/man-pages/man1/false.1.html