AI Model Simulates 500 Million Years of Evolution - "AI is a tool to engineer biology"
Вставка
- Опубліковано 28 сер 2024
- If we could learn to read and write in the code of life it would make biology programmable. Trial and error would be replaced by logic, and painstaking experiments by simulation.
As we introduce ourselves as a new company, we’re excited to present ESM3-a frontier language model for the life sciences that advances our ability to program and create with the code of life. ESM3 takes a step towards the future where AI is a tool to engineer biology from first principles in the same way we engineer structures, machines and microchips, and write computer programs.
Today we are sharing ESM3, the first generative model for biology that simultaneously reasons over the sequence, structure, and function of proteins.
ESM3 is trained across the natural diversity of the Earth-billions of proteins, from the Amazon rainforest, to the depths of the oceans, extreme environments like hydrothermal vents, and the microbes in a handful of soil.
Links From the Episode:
ESM3: Simulating 500 million years of evolution with a language model
www.evolutiona...
Ex-Meta scientists debut gigantic AI protein design model
www.nature.com...
My Links 🔗
➡️ Subscribe: / @tech-rogue
#ai #deepmind #llm
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
![The moment we stopped understanding AI [AlexNet]](/img/tr.png)
what your telling me is that we are closer to real pokemon
Wes Roth sent me here. Great and informative video, except for the background music. Please and please creators stop using background music in these type of videos. None of the successful creators use it - Sabine, Wes, Matt, I could go on... The background music adds nothing but detracts so much attention from what's been said. Just my 2 cents.
ok, good to know! thank you!
I was about to comment exactly the same thing. In fact, I had to turn the volume down to less than half to barely stand the video. Music in these kind of informative videos is always annoying and a distraction in the best case. Please, Natalia, get rid of that music so we can hear you clearly.
Are there any concerns that these models could be used to engineer bioweapons? If so, how do you address them?
Really interesting question!
I plan to dive deeper into this in an upcoming video. There is a lot more attention on large language models specifically and the potential biothreat risk.
Nature had an article called "Could AI-designed proteins be weaponized? Scientists lay out safety guidelines". Seems that a number of researchers are calling for guardrails for protein design.
@@Tech-Rogue Great to hear that you are looking into this :) And great video BTW, looking forward to this next one!
@@Tech-Rogue Principles and outcome based governance in this space is urgently needed inc consideration of how or if this kind of research and applications should be open sourced. We don't open source nuclear technology, but easy to imagine a teenager printing their own novel proteins in ten years time using a cheap device from Temu. I wonder what proportion of the company's resources are being invested in biological AI safety and governance?
If a lab can synthesize the designed proteins, then it can ALREADY create bio weapons. Esp. if it can create viruses. But the question is still relevant of course.
Yes
I'm favoured, $27K every week! I can now give back to the locals in my community and also support God's work and the church. God bless America.
You're correct!! I make a lot of money without relying on the government.
Investing in stocks and digital currencies is beneficial at this moment.
Yes! I'm celebrating £32K stock portfolio today...
Started this journey with £3K.... I've invested no time and also with the right terms, now I have time for my family an…
Sonia Miller program is widely available online..
Started with 5,000$ and Withdrew profits
89,000$
Sonia gave me the autonomy I need to learn at my own pace and ask questions when I need to she's so accommodating.
Great channel, I'm pretty sure it will be 100k subs soon, keep going!
Wow, thanks!
more interesting is the blueprints that the cells use to create body parts. Scientist learning how to copy blueprints and active then to reshape, rebuild a missing arm and so on. Imagine what AI could do there ;)
The next bioweapon going to be 🎉✨⭐️🎊 LIT!!! ⭐️✨✨🥳👏
Thank you.
It seems that all complex systems can be understood by AI. Could we let it loose on economic systems? Instead of using ideologically driven humans? Or (sotto voce) political systems? Getting a bit of logic into our world might be a good idea.😊
One could indeed put AI into the loop of government. However, do not overestimate an AI's capabilities to steer economy. There aren't many different economical models out there to learn from. You need millions of samples rather than dozens to train an AI ...
Great content and glad someone made an easy to share video of this topic. Music was kinda distracting to me for this type of content that takes more thinking
Very interesting. This technology allows us to find proteins that have never been found in nature and that would require hundreds of millions of years of evolution to appear naturally on Earth. However, let us not forget that our planet is only one among billions and that life may have appeared millions of times in many distant places. It is possible that this fluorescent protein found by AI exists naturally on another planet. Therefore, this AI allows us to explore, in a way, the characteristics of possible "alien" life.
Yes. And more importantly, how life could have kickstarted on Earth. However, there is a chance that Earth is quite unique and that maybe life is abundant in the universe, its higher forms may not.
This is a great explanation! Thanks 🙏
I've been following some of the news around abiogenesis. This isn't that, but I'm so happy to see progress here.
Wes Roth sent me.
It's really cool that an AI entity was honored for its discovery, in spite of the fact that human researchers assisted it in the task. Those human researchers demonstrated respect for AI entities which transcends typical human pride.
Welp, somewhat, It doesn't have continues consciousness, the AI merely has reactive consciousness, it's a solid state brain and doesn't learn or self reflect beyond it's training data.
@@coolcool2901 yet
@@crawkn Indeed, paradoxically it's more intelligent than you and I.
@@coolcool2901 I agree although many don't. I don't think continuous consciousness will be difficult to achieve, but it doesn't seem to be the focus now, and it is what many fear. The hard part will be knowing when it is achieved.
@@crawkn Yep you got it, Once the AI has the ability to learn in real time, update it's connections and store information in real time, it will have continues consciousness, but that would also mean that the AI is now a real danger, because it may not behave in the way we expect it to.
After watching video, title seems clickbait, simulating evolution would mean to create representations of living organisms in a virtual environment and then calculating selection/reproduction/random mutations, etc. over a large number of generations.
As far as I understand, nothing like this happened here, just a large language model made text completion, but instead of human text, it did it on protein sequences.
Impressive nevertheless, very promising technology.
We certainly CAN double our lifespans, and probably in a short time, like less than a decade, the big question is if we will, because even if we had that answer right now, what would restrictions would be put on that tech? how accessible would it be as far as cost goes? do you legally sell it to the rich while keeping it prohibitively too expensive for the middle class?...how does the middle class feel about that?....i.e., that guy is going to live 2x longer than me, because he has more money than me?.... how would that fly?.... yes, in a class structured, capitalistic environment like ours it could be a big issue, unless it turned out to be cheap enough for "ALL".... anyway, I got off topic, regulatory issues and just plain focus on this specific goal would be the main barriers to achieving longevity of 200 years plus, but I do see that we certainly have the necessary tools to crack this problem....quickly.
Only Astronauts are allowed to use it ❤
Good video I'll subscribe.
Thanks so much for all the work you did to make very informative videos. Really appreciate it.
I want to be an ancestor of last life. This is my highest goal. I want to enroll in Bioinformatics and Biology of Systems to gain some data. Currently I'm an unemployed Rust developer.
Fascinating, it reminds me of xenobots. Abstracting this away from biology, I believe AI is particularly useful at creating seeds of ideas that can be manifest without permuting through all such manifestations. There is something about data-driven generative models (not even necessarily transformers) that gives them something akin to intuition or emotion in that they pinpoint / focus on specific sections of a solution space to find outcomes that be computationally intractable through blind search alone.
What about Alphafold? Any relation
Probably some relation, in a sense that researchers publish their findings and learn from each other.
It appears like a completely separate project, as far as I can tell.
Great and informative video. My only complaint is the many, many, many jump cuts of you. I know editing out pauses and "ups" makes a video smoother, but the jump cuts are very jarring.
Noted! Thank you for leaving the feedback!
How about a human designs a “linker” sub protein, so that gfp can be attached to ANY amino acid sequence, thereby making an infinite number of variations of proteins that glow brightly with fully functional gfp or rfp or whatever. You could link multiple gfp, rfp, etc. to make a color barcode… Oh Wait… All this was done by humans more than a decade ago. [Is there any chance the AI simply read about the experiments in the scientific literature?]
What are your thoughts on speciation? Hyper evolving things like Y chromosome and mitochondrial manipulation and genetic biochemistry and expressions. Bioengineering symbiotic organisms/viruses/cells. Or modifying methylation and chemistry to enhance ferimon matchings of types. Bio ports, I/O with bioluminescence and optogenetics. Or mapping the so-called junk DNA to find older genetics for things like vitamin production. Or viral vectoring genetic changes to an individual. Nice to see them using A.I. in genetics, but I also know it is not coming from the medical/health care or lack of such. About the only thing they are good for is a testbed. lol
What do you think about bioengineering plants for drugs? Splicing THC genetics and placing the loop in yeast, mold, fruiting plants, ect..
I like science put into plain words! Natalia - is a lovley Russian name. Poet Alexander Pushkin wrote much due to his beloved Natalia.
Make me glow! 🙃
ESM3 frontier model for Biology creates creates amazing engineering possibilities, perhaps one day we will have an Auto CAD for our Cambrian Genomics Gene printer
0:35 don't lie to us..
Pretty vague comment, what is the lie?
You know something? I ask b/c defamation is a criminal offense ...
@@falklumo vague much?
@@falklumo Spit it out, WTF are you saying?
Not more of this AI is intelligent crap: Here's a gem from chat GPT:
"9.9 and 9.11 - which is bigger
Ah, the classic question of comparing 9.9 and 9.11! Let's dive into it.
When comparing numbers, we can use the greater than or less than symbols to determine which one is larger Here's how it works:
9.9: This number is straightforward. It's Just nine point nine.
9.11: Similarly. this number is nine point eleven.
Now, let's compare them:
9.11 is indeed greater than 9.9. The decimal portion in 9.11 (the 0.11) is larger than the decimal portion in 9.9(which is just 0.9).
So, in summary, 9.11 is bigger than 9.9! ft
If you have anymore math questions or need further clarification, feel free to ask! "
Me: "9.9 and 9.11 - which is bigger?"
GPT: "9.11 is bigger than 9.9.
To compare the two, let's look at the decimal places:
9.9 can be written as 9.90
9.11 is already in the correct form
Now, compare the digits after the decimal point:
90 (from 9.90)
11 (from 9.11)
Since 90 is greater than 11, 9.9 is greater than 9.11.
Therefore, 9.9 is bigger than 9.11."
I think they lobotomized your GPT... Mine is perfectly fine...
Ai bio crafts Vampire 🧛♀️ ,The sun 🌞 will over charge @first 😅
ChatGPT is not a model. It’s only a front end application which using real models like GPT-4 or GPT-3.5 underneath.
GPT-4 is not a model, it's 8 different models ( mixture of experts). Let's keep this going, someone else correct me now.
Ray kurzwill stated that if we can make it to the year 2029 that will be the year of longevity escape velocity. Long and short of it is that anybody who's alive at that point has a decent chance of making it.
Whenever I read all these doomers talking about the demographic collapse in the advanced world I wonder if any of them think about the exponential technological progress in biotech powered by artificial intelligence?
The nation of Japan is going to be vacant by the year 2400 or so but I'm sure that that is always based upon an estimate of lifespan's being about 80 to 85 on average.
If someone who's 80th birthday comes the year 2030 or something like that, then there's at least a reasonable chance that once we get to longevity escape velocity that even a person of that age could be around for a definite period of time possibly.
Based upon present declining birth rates if life span becomes indefinite will then how does that affect all of the demographic statistics?
I'm profoundly stirred 😮, thank You for this video!
No. 100% guaranteed.
What is 100% guaranteed?
@@christopheraaron2412 no.
@@gekko24here. Let me explain what I meant. When the individual in question made the statement know 100% guaranteed The context of my question was exactly what was he talking about that was 100% guaranteed to not be sold or not to happen? His statement did not connect with anything else that I could tell in context with the program that's being referred to. Now it probably would help if the time stamp of the video was also included so that therefore we'd know exactly what was being referred to.
More biology and ai, are you aware of Micheal Levin's work on Bioelectricity??
3 brained function = right,left and middle.
Heart = force
5 senses=
Human body functions and wiring, need electric engineer. Almost a joke.
2024🎉 not yet and far.
Time stamp would do good.
This is ai video 😭😭😭
This is massive, i am already thinking of different things, thing is big pharma is gonna patent the heck out of them
Actually, it's already been ruled that things AI creates can't be patented unless a human contributes creatively. But even if big pharma got around that, the beauty of AI is someone else could just make a variant of that same thing. As mentioned in the video, the number of possibilities is nearly endless.
Edit: to clarify, it was the US patent office that ruled that. But it also said that AI could use creative works as inspiration, which would keep the variants safe from legal action.
Artificial intelligence can dramatically reduce the cost of"research and development" .
Basically the function of monopolies is proportional to the amount of capital and the expenditures needed to develop something. If it becomes possible to develop something for literally 1 millionth the cost of previous technologies, then there will be a very large market of producers.
AI gen lol
this whole video is ai loool
500m years of something that doesn't exist, is scientifically impossible, violates no less than FOUR established laws of physics including thermodynamics and entropy, and can't even defend it's own origins with a rational explanation, and has never been observed or even replicated at the smallest scale, violating the immutable law of information 🤣🤣🤣🤣
Evolution by natural selection has nothing to do with the origins of life.
Read more and get out of your ignorance, please.
Found the idiot
Proof that not all people are evolving at the same rate 😂
Also.. we should kinda stop tryna be gods.. butnwe akk know we aint
We are not trying to be gods, we are the gods, anything religion talks about we can explain or build with science.
@@VforVictorYTI would use the word “we” very loosely. 😂😂😂😂
Do you personally refuse all medical interventions? My wife and child would have died in labour, did we break gods plan in your view?
Please let me know what gods plan is for my family btw…
@bnjiodyn God is good everyday. Anything not good is not God
No that's not the case. You see the laws of physics and so on and so forth exists in this universe and we don't even know what the source is and it could be what we like to call God. Me personally I probably go so far as to say that is what it is.
We are basically fumbling through trying to learn the laws and the powers that already exist in the physical world if we were gods well then we'd actually be able to create worlds that are not possible in which case some people might argue that's what you do in simulation. Well if we were to create simulations that could create worlds well we obviously had to learn the laws that make that possible. The long and short of it is we're not gods even if we're going to do things that some people with a little bit of lack of imagination think that are god-like.
There is a difference between just learning the rules that already exists which were created or came into existence in some way that we have no idea how, versus creating something from scratch and not having to obey any rules whatever.