3 ms·

Machine learning is solving more and more of the optimization problems that quantum computing was ,,supposed to'' solve. AlphaFold is a great example as it is g

Machine learning is solving more and more of the optimization problems that quantum computing was ,,supposed to'' solve. AlphaFold is a great example as it is getting better every year at simulating biochemistry. The only important problem set left is beaking cryptosystems which gives governments even greater power over us.

cma 4y ago
On the other hand the governments benefiting most from strong public key cryptography enabling cryptocurrency are dictatorial ones like North Korea using it to launder past sanctions. For private personal communications you can use symmetric or one time pad since you can exchange keys/pads in person and quantum hasn't been shown to put that at risk (and even has enhanced flavors of them that are even more secure).
- xiphias2 4y ago
  North Korea launders most of its money through US banks (billions) compared to tens of millions going through cryptocurrencies. I worked very hard for many years and payed taxes legally and bought Bitcoin 100% legally through regulated exchanges so it's quite boring to me that money laundering comes up all the time. Most of the money that governments steal are made by tax abiding hard working citizens, not dictatorships that destroyed all their wealth.
ThePhysicist 4y ago
Not sure why this is getting downvoted, I think it's a good insight. There's often more than one way to solve a problem, and ML has shown that one can solve certain problems by extracting commonalities from existing solutions, without understanding the true mechanism below. AlphaFold showed that we can "solve" protein folding without solving the underlying quantum mechanics, which is something that can probably be done in other areas as well. That said there might be some problems that might only be solvable using a full quantum simulation, though personally I think that class of problems will be rather small.
Arkhaine_kupo 4y ago
If I am not wrong, some of the probelms AlphaFold can solve are quite limited. Even if the results are promising, protein folding is an insanely complicated problem and having tools like alphaFold be good at some aspects of it can be useful but cannot crack the bigger nut by design as far as I am aware. It currently performs ok on protein prediction but cannot do new proteins, which would theoretically be possible with quantum computers. And its not just a matter of giving alphafold more resources they are entirely different problems
- xiphias2 4y ago
  I don't totally understand your point. AlphaFold is great at protein folding, and the main limitation is predicting protein structures with multiple states when protein interactions happen. That's the next task the team is working on (it's the natural next step for biology simulation), and if the team is successful, it will be incredibly valuable (it can easily be the next multibillion dollar business for Alphabet, and a huge improvement in how drug research is done. I don't see quantum computers anywhere near as close to solving a real problem with drug research as AlphaFold (although I don't argue with the theoritical possibility in the future, but not in the next 5-10 years, where AlphaFold can shine).
  - Arkhaine_kupo 4y ago
    I apologise as my field is not related to proteins, but I do recall an interesting read about the subject from a few months ago. Here is the link if you wanna follow it yourself. https://news.ycombinator.com/item?id=32262856 https://news.ycombinator.com/item?id=32262856 There are some interesting criticisms from the language, to the results, to the limitations of AlphaFold that I think seem to represent the prevailing opinion from biologists about the current scope of what ML can achieve on their field. There is also push back to the complaints and some very optimistic people about Deep learning models on the field so perhaps you are right.
hulitu 4y ago
> AlphaFold is a great example as it is getting better every year at simulating biochemistry. So was also seti@home.

cma 4y ago

On the other hand the governments benefiting most from strong public key cryptography enabling cryptocurrency are dictatorial ones like North Korea using it to launder past sanctions. For private personal communications you can use symmetric or one time pad since you can exchange keys/pads in person and quantum hasn't been shown to put that at risk (and even has enhanced flavors of them that are even more secure).

xiphias2 4y ago
North Korea launders most of its money through US banks (billions) compared to tens of millions going through cryptocurrencies. I worked very hard for many years and payed taxes legally and bought Bitcoin 100% legally through regulated exchanges so it's quite boring to me that money laundering comes up all the time. Most of the money that governments steal are made by tax abiding hard working citizens, not dictatorships that destroyed all their wealth.

ThePhysicist 4y ago

Not sure why this is getting downvoted, I think it's a good insight. There's often more than one way to solve a problem, and ML has shown that one can solve certain problems by extracting commonalities from existing solutions, without understanding the true mechanism below. AlphaFold showed that we can "solve" protein folding without solving the underlying quantum mechanics, which is something that can probably be done in other areas as well. That said there might be some problems that might only be solvable using a full quantum simulation, though personally I think that class of problems will be rather small.

Arkhaine_kupo 4y ago

If I am not wrong, some of the probelms AlphaFold can solve are quite limited. Even if the results are promising, protein folding is an insanely complicated problem and having tools like alphaFold be good at some aspects of it can be useful but cannot crack the bigger nut by design as far as I am aware. It currently performs ok on protein prediction but cannot do new proteins, which would theoretically be possible with quantum computers. And its not just a matter of giving alphafold more resources they are entirely different problems

xiphias2 4y ago
I don't totally understand your point. AlphaFold is great at protein folding, and the main limitation is predicting protein structures with multiple states when protein interactions happen. That's the next task the team is working on (it's the natural next step for biology simulation), and if the team is successful, it will be incredibly valuable (it can easily be the next multibillion dollar business for Alphabet, and a huge improvement in how drug research is done. I don't see quantum computers anywhere near as close to solving a real problem with drug research as AlphaFold (although I don't argue with the theoritical possibility in the future, but not in the next 5-10 years, where AlphaFold can shine).
- Arkhaine_kupo 4y ago
  I apologise as my field is not related to proteins, but I do recall an interesting read about the subject from a few months ago. Here is the link if you wanna follow it yourself. https://news.ycombinator.com/item?id=32262856 https://news.ycombinator.com/item?id=32262856 There are some interesting criticisms from the language, to the results, to the limitations of AlphaFold that I think seem to represent the prevailing opinion from biologists about the current scope of what ML can achieve on their field. There is also push back to the complaints and some very optimistic people about Deep learning models on the field so perhaps you are right.

hulitu 4y ago

> AlphaFold is a great example as it is getting better every year at simulating biochemistry. So was also seti@home.