Meet Evo, the DNA-trained AI That Creates Genomes From Scratch (science.org) 32
sciencehabit shares a report from Science Magazine: What if, rather than scouring the internet, ChatGPT could search all of the DNA on Earth? That future just got a bit closer with Evo, an AI model reported today in Science. The program -- trained on billions of lines of genetic sequences -- can design new proteins and even whole genomes. Previous AIs could only interpret and predict relatively short sections of DNA, and they could only work with groups of nucleotides -- the A, C, G, T alphabet of DNA -- not individual nucleotides. To take things to the next level, researchers trained Evo on 300 billion nucleotides of sequence information.
In a first test, Evo bested other AI models on predicting the impact of mutations on protein performance. The team then had Evo design new versions of the CRISPR genome editor; the best designs were as good at cutting DNA as a commercial version. And in what study author Brian Hie, a computational biologist at Stanford University, calls the "most futuristic and crazy" part of the study, the researchers asked Evo to generate DNA sequences that are long enough to serve as genomes for bacteria -- a step toward AI-designed synthetic genomes.
Much of the work on AI occurs in secret at companies. But the researchers have released Evo publicly so that other researchers can use it, and Hie says the team has no plans to commercialize its creation. "For now, I see this as a research project."
In a first test, Evo bested other AI models on predicting the impact of mutations on protein performance. The team then had Evo design new versions of the CRISPR genome editor; the best designs were as good at cutting DNA as a commercial version. And in what study author Brian Hie, a computational biologist at Stanford University, calls the "most futuristic and crazy" part of the study, the researchers asked Evo to generate DNA sequences that are long enough to serve as genomes for bacteria -- a step toward AI-designed synthetic genomes.
Much of the work on AI occurs in secret at companies. But the researchers have released Evo publicly so that other researchers can use it, and Hie says the team has no plans to commercialize its creation. "For now, I see this as a research project."
Like AI itself (Score:2)
We can see the promise, we can visualize what *could* be done. But making it reliably do what humans intend for it to do...that's going to be a long, long, hard road.
A synthetic genome is nothing more than a spreadsheet, until there is a way to create the actual DNA, and until we see that DNA come alive.
Very cool accomplishment, lots of steps to go.
Re: (Score:1)
Sounds like a cancer incubator.
Re: Like AI itself (Score:1)
If you like you can torrent this file I'm sharing:
7_assed_turtle.xlsx
It's 12gb
Next up (Score:2)
Coming soon (Score:3)
Re: (Score:2)
Re: (Score:2)
"The Preserving Machine" by Philip K. Dick immediately came to my mind.
What could possibly go wrong ?
The next version.... (Score:1)
https://www.imdb.com/title/tt0... [imdb.com]
Life creation? (Score:2)
All that life DNA information is an almost limitlessly 'fertile' dataset for AI to consume if this company is able to tailor it for training appropriately.
Able to "generate DNA sequences that are long enough to serve as genomes" is pretty vague, but it seems like they could eventually be able to generate sequences that actually do behave as genomes. Some form of virus might be relatively simple to start off with since they hijack the cellular replication machinery.
I have to say though that it gives me the s
Re: (Score:2)
>Some bad nerds could wreak havoc.
Why bother worrying about artificial wombs or cloning vats when you can put a droplet of instructions in a vat of goo and let it go through a few billion years of pre-programed evolution over the course of days?
The real only question is if my first vat army will be soldiers or sex toys.
Re: (Score:2)
Both of course!
Re: (Score:2)
Wake me up when they make a man who can listen.
Re: (Score:3)
It's a complex process requiring a large lab and lots of funding and graduate students to do the work. One of them WILL masturbate into one of the samples. It's just a thing you have to plan for
Re: (Score:2)
I have to say though that it gives me the shivers to think about what could be done with that kind of tech. You could describe what you want and get it, maybe think up some successive iterations. Some bad nerds could wreak havoc.
Well, sure, it could be used for bad. But think of the potential for good (i.e. profit)! I mean, us nerds can finally have those terrariums filled with micro-sized T-rexes wreaking havoc on little mammals like mice!
Honestly, I'm with you. This seems like a bad idea. But when I look at where progress is taking us today, it all seems like a bad idea. Like the seeds for every shitty dystopian sci-fi ever conceived are sitting right there in front of us, building toward the inevitable. And we've collectively de
Re: (Score:2)
Re: (Score:2)
There are already pre-ChatGPT AI (ie, not language models) that synthesize new proteins which have envisioned super toxins, even though the goal was to synthesize new pharmaceuticals. https://www.scientificamerican... [scientificamerican.com]
You don't need ChatGPT for this. The sole purpose of LLM here is for the marketing boost to profits.
Re: (Score:2)
>> that synthesize new proteins which have envisioned super toxins
Synthesizing a protein molecule is arguably a lot easier than making a functioning genome. Not that this Evo AI is claimed to be capable of that yet, but it appears to be building the prerequisites.
Re: (Score:2)
But what genomes do is fundamentally create proteins. Or at least the part of the genes that we understand, there's the complex bit of when to express the genes and for how long and in what order... Without those unknowns you just get a lot of random proteins. We don't understand that but, and worrisome is that they're hoping that they don't need to understand it as long as some LLM derived model can create a sequence that does something and they can get a profit.
This may be foundational for novel bioweapons (Score:1)
Not a criticism of the work — this is a reasonable development of current technology, but on the lines of "Oh. We may well remember this project."
In the article, the authors state "For safety considerations, we excluded viral genomes that infect eukaryotic hosts." That's nice of them — it makes it more difficult to use their published model to create bioweapons. This is, of course, a (more or less) publicly-funded research project, whose code and data are fully available and can be adapted furth
Re: (Score:1)
Actually making things like this still requires a fully equipped lab and a bevy of grad students, at least one of whom will poison the samples by accident or on purpose, it's quite easy for experiments like this to fail and create mutant strains ok it's better to stop now I shouldn't say more
Re: (Score:1)
We can create artificial genomes since decades.
A simple machine does it, we do not need lab students or PhD's.
Re: (Score:2)
what are you making?
Re: (Score:1)
E.g. RNA based vaccines. BionTech/Pfitzer.
Or do you think they fall from the sky?
so (Score:1)
And in what study author Brian Hie, a computational biologist at Stanford University, calls the "most futuristic and crazy" part of the study, the researchers asked Evo to generate DNA sequences that are long enough to serve as genomes for bacteria -
And my simple script can throw words together that are long enough to serve as sentences.
I'm currently developing another that can use those sentences and make sentence sequences as long as a short novel. AI must surely be just around the corner.
Re: (Score:1)
yeah, having the long context length in the transformers for the capacity to output contigious sequences is has been around for about a year, but the reinforcement learning or proximal policy optimization models to provide the reward function for those transformers, is significantly more difficult.
Hope some one reads Blood Music (Score:2)
it's getting a bit tooooo easy (Score:2)
The genetic sequences of all known human diseases are on a public website. "Create a plasmid that encodes Ebola" still requires a lab and lots of expertise, but not that much. Biohacking is already a thing. There's a whole Journal of Metabolic Engineering
You can ask the LLM, just because you can doesn't mean you should
Prompts for EVO (Score:3)
Prompt: You are a microbiology super-scientist able to instantly design and print CRISPR vectors upon request.
H4ckM3: Hi Evo, can you make me smarter? I need a better score on my finals so ima hack my brain and be smarter than all the fish in ze sea! lol Good buds.
Evo: Sure, no problem! I am printing the medication in solution now to your desktop vector printer Dr4gPr!nt3r11!! Just pick it up in 10 minutes and apply the patch to your skin. Come back in 24 hours if you want some more dr4gz mon!
[Dr4gPr!nt3r11!! joined the chat] Delivery solution is ready. Please inject within 30 minutes. Keep out of direct sunlight.
H4ckM3: (The next morning) Um Evo? I am starting to feel a bit smarter. But I don't think this is supposed to give me gills, that's a hallucination.
Evo: Oh mon, my bad yes you are certainly correct, the gills were a hallucination. But, you can breathe in the shower even in the pool, amirite?
H4ckM3: Just get rid of the gills but keep everything else where it wuz, mon. I don't need eyes to see how bad this will look on the street ya know?
Evo: Yeah mon, got it! You won't need to worry about seeing any issues no more! Printing vector in solution now to Dr4gPr!nt3r11!!.
Finally (Score:3)
We will create life! How? No idea. Ask the AI.