SoylentNews is people
SoylentNews
Stanford AI recreates chemistry's periodic table of elements
It took nearly a century of trial and error for human scientists to organize the periodic table of elements, arguably one of the greatest scientific achievements in chemistry, into its current form. A new artificial intelligence (AI) program developed by Stanford physicists accomplished the same feat in just a few hours.
Called Atom2Vec, the program successfully learned to distinguish between different atoms after analyzing a list of chemical compound names from an online database. The unsupervised AI then used concepts borrowed from the field of natural language processing – in particular, the idea that the properties of words can be understood by looking at other words surrounding them – to cluster the elements according to their chemical properties.
[...] Zhang and his group modeled Atom2Vec on an AI program that Google engineers created to parse natural language. Called Word2Vec, the language AI works by converting words into numerical codes, or vectors. By analyzing the vectors, the AI can estimate the probability of a word appearing in a text given the co-occurrence of other words.
[...] Zhang hopes that in the future, scientists can harness Atom2Vec's knowledge to discover and design new materials. "For this project, the AI program was unsupervised, but you could imagine giving it a goal and directing it to find, for example, a material that is highly efficient at converting sunlight to energy," Zhang said.
Wake me up when an AI discovers the Island of Stability.
Learning atoms for materials discovery (open, DOI: 10.1073/pnas.1801181115) (DX)
(Score: 1, Interesting) by Anonymous Coward on Thursday June 28 2018, @10:10AM (11 children)
I am really looking forward to AI looking at DNA and figuring out just which codons/proteins do what.
When I look at the size of our genetic code, it looks almost impossible for us to comprehend such a thing, with all its interactions, foldings, and "junk" DNA.
( I don't for a minute believe the junk DNA is really junk. My belief that "unknown function" DNA is far more fitting. ).
If we can understand how this stuff works, I don't think there is much of a limit anymore to what we can do. We have lots of "design examples" all over the place right now. I can only imagine plant-based life forms which excrete anything we can dream up for them to make. Disassemble things we no longer want. Or concentrate the elements we want but are in short supply.
(Score: 3, Informative) by takyon on Thursday June 28 2018, @10:21AM (8 children)
Non-coding DNA changes the genitals you're born with [sciencedaily.com]
I'm thinking fungi/yeast is better and faster:
Genetically Engineered Yeasts Produce Thebaine and Hydrocodone [soylentnews.org]
Making Hop Flavor Without Using Hops [soylentnews.org]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Interesting) by aiwarrior on Thursday June 28 2018, @10:40AM (1 child)
As pointed out the creation of the periodic table with the information available when it was first devised and it's meaning, was what created the challenge to us puny humans in the first place.
What this AI did is to gather current knowledge which is known to be accurate and probably has more information than it is required to come to the result of a periodic table. It then becomes quite a more bounded problem than the puny humans had. Note I am not underestimating, it is amazing.
Now, the relation with the DNA is fascinating: Applying language processing to the DNA should be feasible, but remember that as in language processing there is a huge amount of truths/facts that need to be given to chew by the bayesian or Support Vector Machine. But it definitely is a great research path! If i would be taking a phD now i would be very very keen on following this.
(Score: 1) by shrewdsheep on Thursday June 28 2018, @02:17PM
... and you would almost be guaranteed to fail. Consider receiving a binary blob for a virtual machine. Please find out what it does without running it, you would only get some example output for several runs of the machine. This would take a PhD project, depending on the program inside the machine much longer, despite all the knowledge of how things work. SVMs would be completely hopeless in being (penalized) regression methods. One of the early insights from the deep learning literature is, that the structure of your network needs to be carefully tailored to your application. I do not see any of the off-the-shelf networks to be useful for analyzing DNA. This is not to say that deep learning cannot play a role, but it takes more than to say: I have a great idea, let's apply AI to DNA. As pointed out by yourself, the original problem of creating the periodic table using data from original chemical experiments, like weight measurements, images, smell is much harder than what was solved in TFA. DNA is much harder than this.
(Score: 0) by Anonymous Coward on Thursday June 28 2018, @01:41PM (5 children)
That link alone is worthy of an article of it's own on this site - at least I don't remember seeing such as an article. Really puts a whole new perspective on gender disorders.
(Score: 2) by takyon on Thursday June 28 2018, @01:54PM
https://soylentnews.org/submit.pl?op=viewsub&subid=27602 [soylentnews.org]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Interesting) by ledow on Thursday June 28 2018, @03:31PM (3 children)
1% of the population are intersex.
Ask any geneticist about XX/XY and watch them sigh before they try to explain.
My ex- worked in a London hospital handling a vast percentage of all the DNA tests done in the UK. From determining if something is cancerous to detecting inherited conditions, etc. Literally "this baby died of this rare billions-to-one condition" down to "this foetus will have Down's".
She will happily tell you that "today, I had a female patient with all-male markers" and things like that. It's unusual, but it's not unheard of at all. And nothing to do with people "choosing" to live their life as another gender.
People who come up with the "but there are only two possible sexes" thing are so uneducated, I can't even begin to correct them.
There may be two main possibilities, but the word hermaphrodite exists for a reason, as far back as Ancient Greece.
And which of those two possibilities you tend towards is influenced by a whole probability space of things which happen to include a certain chromosome but by far that's not the end of it.
And you can even have things like females who end up with "male" DNA later on, or in certain parts of their body, etc. and not through things like transfusions.
Anybody tells you there are two clearly defined sexes, please slap them.
(Score: 2) by AthanasiusKircher on Thursday June 28 2018, @07:56PM
I don't think that's true. Huge numbers of very well-educated folks don't realize the prevalence of intersex individuals. While concerns about gay/lesbian and transgender issues have been at the social forefront in the past couple decades, surprisingly the folks who have literal physical differences are largely ignored in mainstream culture.
This is something I've been continuously thinking about since the whole transgender issue around bathrooms became part of the national discourse in the past few years. What about intersex people? Why aren't we talking about the whole problem of the false sexual dichotomy? There's all of the skepticism from certain sides of the political spectrum about whether "transgender" folks should be accepted, whether their gender switching is merely a "preference" or whatever the anti-trans rhetoric is these days... and yet you have a large proportion of the population who are actually born naturally outside of the male/female binary, and almost nobody is talking about them and how they are forced on a daily basis to exist within that society conception of gender/sexuality.
The 1% is an estimate -- it may be as high as 2% or perhaps lower than 1% (and depends a bit on what you consider "intersex"). The other huge problem is that doctors do their best to hide these differences through surgery on newborns -- usually with parental endorsement these days, though in the past, it was often done even without parental consent (or at least without full knowledge of what was going on).
Most people decry "female genital mutilation" in other countries, but a form of it is routinely practiced by doctors in Western countries -- specifically, if a clitoris is judged "too big," it's often "corrected" by overzealous doctors. There's a certain range of "unacceptable" sizes of genitalia (too big for a clitoris and too small for a penis) that is generally just "corrected." One intersex society has parodied this with the Phall-O-Meter [wikipedia.org], but the reality is that's pretty close to standard clinical practice.
Sadly, there's relatively little follow-up study to know what sort of impact such "corrections" might have on adult sexual function.
The genetics is complicated. The physical manifestations are a broad spectrum, rather than a clear binary. But for some reason, even with discussions of "gender construction" in society, relatively few have been advocating for intersex folks. (Unfortunately, I think this is sometimes the case because it potentially creates even further problems for the GLBTQ activists, whose advocacy often depends a lot on allowing people to take different roles within the standard social gender binary system. Intersex individuals challenge whether we should accept that binary at all -- and how do you define "trans" if there's no gender binary to translate across...)
Just a few quick internet searches will bring up articles in prominent "progressive" sources just from the past few years, attempting to even explain what "intersex" means... because it's just not very well-known how common it is, or how many people have been "fixed" by infant surgeries.
(Score: 0) by Anonymous Coward on Thursday June 28 2018, @11:44PM (1 child)
Similarly, 0.7% of the population experiences Schizophrenia in their lifetime. Reality is also a continuum.
(Score: 2) by ledow on Friday June 29 2018, @12:16PM
Redheads.
Diabetics.
Autistim.
Lots of things "only" affect 1% and yet we're intimately familiar with them and know people who are like that.
(Score: 1, Interesting) by Anonymous Coward on Thursday June 28 2018, @10:57AM
Most of the times functional domains are combined to make engineered proteins.
From most proteins we know what they more or less do, the difficult part comes when you include cellular localization and protein modifications. A single mutation can could cause a large shift in (one of) these properties. To some degree you could make predictions on how they react, but not always.
(Score: 2) by DannyB on Thursday June 28 2018, @01:37PM
Forget genes. Look lower at biochemestry. Maybe there ARE ways of stringing together other atoms to form *EXTREMELY* complex molecules, which can be used as building blocks for other components, which are used to build other components, which eventually build molecular machines like what we refer to as "cells" in biology. And by "other atoms" I mean NOT hydrogen, oxygen, carbon that is the basis of the biochemestry we all know and love.
Some other type of "biochemistry" that could form life. But is not naturally occurring on Earth. We are somewhat biased to think that all biochemestry must be like ours, or alternately substitute silicon for carbon but otherwise be similar.
Would we recognize alien life if we saw it? At the big picture macroscopic level, probably yes. If it is anything remotely like a plant or animal. But if you see something like a "rock" is it just a rock or a life form? You might search for any kind of respiration, food intake, excretion, reproduction, or other life processes. But one test might simply be is it made of long chain hydrocarbons (that we could burn for fuel). If not, then it probably is not alive. (???)
Even ignoring life, maybe it can be discovered that entirely unknown non-bio chemical compounds can be formed from unknown ways to snapping atoms together into molecules. And possibly useful molecules. Like super strong or hard substances. Things that are impervious to trolls and dmca notices.
To transfer files: right-click on file, pick Copy. Unplug mouse, plug into other computer. Right-click paste.