[DocSavage87]

This is more on the basic science side, but I'd expect it to have impacts throughout biology and medicine (including vaccine and therapeutics development) in the years to come.  As little good news as we've had this year, we should celebrate when human ingenuity makes real breakthroughs.  The field of deep learning is still in its relative infancy so I'd expect the technology to continue to improve as both algorithms and hardware improve.

Here's a tweet thread summarizing it with links to the DeepMind blog post and the Nature news article:

[tweet]
https://twitter.com/MoAlQuraishi/status/...13280?s=20
[/tweet]

[DocSavage87]

This is more on the basic science side, but I'd expect it to have impacts throughout biology and medicine (including vaccine and therapeutics development) in the years to come.  As little good news as we've had this year, we should celebrate when human ingenuity makes real breakthroughs.  The field of deep learning is still in its relative infancy so I'd expect the technology to continue to improve as both algorithms and hardware improve.

Here's a tweet thread summarizing it with links to the DeepMind blog post and the Nature news article:

[tweet]
https://twitter.com/MoAlQuraishi/status/...13280?s=20
[/tweet]

[DocSavage87]

It's amazing how much better the new Deep Mind approach is to protein folding.  Take a look at this chart of the various submitted solutions to the protein folding competition.  AI / Deep Learning hopefully has a big of an impact on biology as it has on computer vision.



To temper my enthusiasm, there are some group leaders who say this is great but it doesn't mean we can do structure guided design easily yet, although there are some bets between labs already in motion.  The level of improvement on a fundamental problem in basic science is great and bodes well for faster biomedical R&D.

[DocSavage87]

It's amazing how much better the new Deep Mind approach is to protein folding.  Take a look at this chart of the various submitted solutions to the protein folding competition.  AI / Deep Learning hopefully has a big of an impact on biology as it has on computer vision.



To temper my enthusiasm, there are some group leaders who say this is great but it doesn't mean we can do structure guided design easily yet, although there are some bets between labs already in motion.  The level of improvement on a fundamental problem in basic science is great and bodes well for faster biomedical R&D.

[BostonCard]

Thanks for linking.  This is certainly exciting, if it pans out (obviously, a ways to go before we see its widespread use, rather than individual cases).  And for deep mind, which has been criticized for using its AI to master chess and go, it is a good retort.

BC

[BostonCard]

Thanks for linking.  This is certainly exciting, if it pans out (obviously, a ways to go before we see its widespread use, rather than individual cases).  And for deep mind, which has been criticized for using its AI to master chess and go, it is a good retort.

BC

[Snorlax94]

Wow, thank you for sharing this! I think it's brobdingnagian -- IMO, this is the biggest news of the week, and possibly bigger.

There was a NYTimes article about Deep Mind two years ago where its machine-learning algorithm AlphaGo, after learning the rules of Go, and no further assistance, could play itself millions of times and in a matter of hours, become better than the best player in the world. That article was actually the Holy SH**T moment for me where I could start to see speculative science fiction scenarios -- possibly for good, possibly for bad -- becoming reality in our lifetimes.

It's amazing to think that less than two years later, it is making breakthroughs in biological or medical science. The biology/medicine is outside of my areas of study, but from my limited knowledge from organic chemistry (chem 33 and 35?), it sounds like Deep Mind is taking the chemical formula for very large and complex molecules (proteins) and deducing its geometric structure based on known properties. An NP-complete problem that Deep Mind is able to solve comparably well to humans, but in hours instead of years or decades.

One researcher said he already used the algorithm to solve a previously-unsolved chemical formula. So, it seems the implications and adoption are already at hand.

It is exciting and wonderful but also a little bit scary. Aside from the "Skynet" (the sentient AI that declares war on mankind from the Terminator movies) type of sci-fi scenarios that may be unlikely (though worth taking caution against), I'm also very concerned about how much power can be created and concentrated from this technology. Unlike Skynet, a massive concentration of wealth and power, and the ability to impact humans in ways many people won't be able to grasp, is a very possible outcome of AI. From the Lord of the Rings, I call AI "the one technology to rule them all."  Deep Mind's breakthrough in protein folding gives an early example of how and why whoever has the best AI may end up having the best technology in many fields.

[Snorlax94]

Wow, thank you for sharing this! I think it's brobdingnagian -- IMO, this is the biggest news of the week, and possibly bigger.

There was a NYTimes article about Deep Mind two years ago where its machine-learning algorithm AlphaGo, after learning the rules of Go, and no further assistance, could play itself millions of times and in a matter of hours, become better than the best player in the world. That article was actually the Holy SH**T moment for me where I could start to see speculative science fiction scenarios -- possibly for good, possibly for bad -- becoming reality in our lifetimes.

It's amazing to think that less than two years later, it is making breakthroughs in biological or medical science. The biology/medicine is outside of my areas of study, but from my limited knowledge from organic chemistry (chem 33 and 35?), it sounds like Deep Mind is taking the chemical formula for very large and complex molecules (proteins) and deducing its geometric structure based on known properties. An NP-complete problem that Deep Mind is able to solve comparably well to humans, but in hours instead of years or decades.

One researcher said he already used the algorithm to solve a previously-unsolved chemical formula. So, it seems the implications and adoption are already at hand.

It is exciting and wonderful but also a little bit scary. Aside from the "Skynet" (the sentient AI that declares war on mankind from the Terminator movies) type of sci-fi scenarios that may be unlikely (though worth taking caution against), I'm also very concerned about how much power can be created and concentrated from this technology. Unlike Skynet, a massive concentration of wealth and power, and the ability to impact humans in ways many people won't be able to grasp, is a very possible outcome of AI. From the Lord of the Rings, I call AI "the one technology to rule them all."  Deep Mind's breakthrough in protein folding gives an early example of how and why whoever has the best AI may end up having the best technology in many fields.

[JohnR34231]

For those interested here is an article that talks about it.

https://www.statnews.com/2020/11/30/deep...-150622929

[JohnR34231]

For those interested here is an article that talks about it.

https://www.statnews.com/2020/11/30/deep...-150622929

[Mick]

It is exciting and wonderful but also a little bit scary. Aside from the "Skynet" (the sentient AI that declares war on mankind from the Terminator movies) type of sci-fi scenarios that may be unlikely (though worth taking caution against), I'm also very concerned about how much power can be created and concentrated from this technology. Unlike Skynet, a massive concentration of wealth and power, and the ability to impact humans in ways many people won't be able to grasp, is a very possible outcome of AI. From the Lord of the Rings, I call AI "the one technology to rule them all."  Deep Mind's breakthrough in protein folding gives an early example of how and why whoever has the best AI may end up having the best technology in many fields.

It's just a little thing, but a friend of mine spent a year devising an AI approach to horse race handicapping and is doing very, very well.

[Mick]

It is exciting and wonderful but also a little bit scary. Aside from the "Skynet" (the sentient AI that declares war on mankind from the Terminator movies) type of sci-fi scenarios that may be unlikely (though worth taking caution against), I'm also very concerned about how much power can be created and concentrated from this technology. Unlike Skynet, a massive concentration of wealth and power, and the ability to impact humans in ways many people won't be able to grasp, is a very possible outcome of AI. From the Lord of the Rings, I call AI "the one technology to rule them all."  Deep Mind's breakthrough in protein folding gives an early example of how and why whoever has the best AI may end up having the best technology in many fields.

It's just a little thing, but a friend of mine spent a year devising an AI approach to horse race handicapping and is doing very, very well.