Slashdot Mirror
AI Tool, Which Has Digested Nearly Every Reaction Ever Performed, Can Invent New Ways To Create Complex Molecules (nature.com)
An anonymous reader shares a research paper: Researchers have developed a 'deep learning' computer program that produces blueprints for the sequences of reactions needed to create small organic molecules, such as drug compounds. The pathways that the tool suggests look just as good on paper as those devised by human chemists. The tool is not the first software to wield AI instead of human skill and intuition. Yet chemists hail the development as a milestone, saying that it could speed up the process of drug discovery and make organic chemistry more efficient. "What we have seen here is that this kind of artificial intelligence can capture this expert knowledge," says Pablo Carbonell, who designs synthesis-predicting tools at the University of Manchester, UK, and was not involved in the work. He describes the effort as "a landmark paper."
[...] Chemists have conventionally scoured lists of reactions recorded by others, and drawn on their own intuition to work out a step-by-step pathway to make a particular compound. They usually work backwards, starting with the molecule they want to create and then analysing which readily available reagents and sequences of reactions could be used to synthesize it -- a process known as retrosynthesis, which can take hours or even days of planning. The new AI tool, developed by Marwin Segler, an organic chemist and artificial-intelligence researcher at the University of Munster in Germany, and his colleagues, uses deep-learning neural networks to imbibe essentially all known single-step organic-chemistry reactions -- about 12.4 million of them. This enables it to predict the chemical reactions that can be used in any single step. The tool repeatedly applies these neural networks in planning a multi-step synthesis, deconstructing the desired molecule until it ends up with the available starting reagents.
[...] Chemists have conventionally scoured lists of reactions recorded by others, and drawn on their own intuition to work out a step-by-step pathway to make a particular compound. They usually work backwards, starting with the molecule they want to create and then analysing which readily available reagents and sequences of reactions could be used to synthesize it -- a process known as retrosynthesis, which can take hours or even days of planning. The new AI tool, developed by Marwin Segler, an organic chemist and artificial-intelligence researcher at the University of Munster in Germany, and his colleagues, uses deep-learning neural networks to imbibe essentially all known single-step organic-chemistry reactions -- about 12.4 million of them. This enables it to predict the chemical reactions that can be used in any single step. The tool repeatedly applies these neural networks in planning a multi-step synthesis, deconstructing the desired molecule until it ends up with the available starting reagents.
I'm waiting for Slashdot to implement AI. To prevent duplicate posts like these.
https://science.slashdot.org/story/18/03/28/232206/new-deep-learning-software-knows-how-to-make-desired-organic-molecules
How about an AI tool that has digested every Slashdot article posted less than 48 hours ago so it doesn't get posted again? We can name it "Ed".
But... why does it keep calling itself "Skynet"?
I've abandoned my search for truth; now I'm just looking for some useful delusions.
Don't make molly, make Molly!
I've abandoned my search for truth; now I'm just looking for some useful delusions.
If only we had the ability to come up with an AI that would identify duplicate slashdot stories....
"If it ain't broke, it doesn't have enough features yet"
You expect AI to be more than a set of programming instructions?
Here is the article which the Nature summary linked in the ./ summary summarizes.
The Nature summary links to that article and states "The new AI tool, developed by Marwin Segler, an organic chemist and artificial-intelligence researcher at the University of Münster in Germany..." Weirdly, Segler is not listed as an author on the article and none of the article's authors are at Münster. Even stranger, Segler is not even cited.
Ceci n'est pas une signature.
What it's doing is a deep learning version of what can be done by a pruning branch analysis by working the reactions in reverse. The sections of indecision in reactions are replaced by the AI decision formed from data dumps as opposed to some programmed heuristic.
It's not a huge leap in AI or VI or whatever someone wants to call it these days, but it's a good application of 'deep learning'.
I don't read AC
Queue up the guy that will say it isn't AI because it isn't Lt Cmdr Data! Anyhow, this sort of modeling sounds like it could be pretty useful.
Assuming that "AI" isn't just the new word for "any computer program that sounds complicated to laymen" then I'm guessing it's because the for loop wouldn't finish in your life time. So they tried out all the usual tricks that people use to avoid combinatorial explosion, the various other tricks for optimizing functions that you can't handle with just plain old math, and finally at the bottom of their bag of tricks, there was AI -- and it worked!
"Believe me!" -- Donald Trump
ObXKCD
"Believe me!" -- Donald Trump
Great. Now we can really fix things.
The Russians have won. They have made the world a cesspool of distrust, greed, fear and hate.
It isn't. It's just programming. Doesn't matter how the data got in there, it's just data being executed upon by software like any other data being executed upon. This is Slashdot though, so OMG AI.
Perhaps this could be used to discover new Hypothetical types of biochemistry and we can then speculate about many more types of alternative biochemistries ("life as we don't know it").
Anybody up for Nano Tube production?