Why isn't openAI used in biology/medecine?
I believe the potential would be much greater there, instead of scrapping content from internet, it would try to figure out how human cells and body work by simulating embryonic cells.
not only it would help cure most diseases and potentially aging it would also remove the need for animal testing.
Mystery of the ages.
Is only game! Why heff to be mad?
You don't know how anything works is the problem. Plenty of simulations and systems exist to attempt predicting and modeling the human body, same goes for innumerable attempts to construct successful assistance tools and machine learning models. "OpenAI" is just modern bonzai buddy and way less impressive than you seem to think it is. You'd just get 2023 bonzai buddy saying everything is lupus and cancer.
are you telling me that the current gpu super computers can't simulate a single embryonic cell?
It can't. Nothing currently can simulate the complex mechanics and molecular interactions of a single cell, nevermind something as complex as a plant
GPT doesn't compute moron. It can generate text for you. If you prompt it in a certain way that it's text data to text data it can be useful. But for biological simulation there are usually more than 4000 tokens of data to handle, and different associations between them.
The news is fear mongering. Gpt 4 5 6 7 8 etc will never compute, because it's simply not built to.
The AGI nonsense is psuedoscience.
>usually more than 4000 tokens of data to handle
sounds like a job for hyena
>https://archive DOT ph/3XLhQ
>From Deep to Long Learning?
>Plus, the convolution can be computed in O(N log N) time in sequence length – nearly-linear scaling!
>With some optimizations (more on that below), Hyena models are slightly slower than Transformers of the same size at sequence length 2K – but get a lot faster at longer sequence lengths.
>are you telling me that the current gpu super computers can't simulate a single embryonic cell?
Depends what you mean by simulate. You could “simulate” a cell by modeling it as an oval if you want, but you won’t get much information from that. If you want to actually simulate all of the internal workings at a quantum level, no nothing is even close at the moment.
You don't necessarily need quantum-level fidelity to have useful models. Simulations have been made that predict cell behavior of various sorts with fair accuracy, even if a complete life simulation still seems out of reach.
We can model every aspect of a car, at least, without the need for quantum level simulations, but of course, cars are considerably more simple than cells.
It can’t simulate cells that. It’s a language model. You’d need a super computer and a different simulation system.
>You’d need a super computer
probably
>and a different simulation system.
anon, I...
LLMs are nothing but autocomplete technically. Predict the next word over and over again. Yet they now take and follow instructions. Despite no specific code written there are features that arise latent in the models as parameter and layer sizes grow.
Even if LLMs don't get used directly they could have a place as tools or assistants to speed up the work of humans. Here's a look at chemistry for example
https://arxiv.org/abs/2304.05376
A collection of unintegrated tools can be integrated for use with a LLM.
Last year protein folding was solved by an AI developed by Google. It discovered billions of previously unknown proteins, in particular thousands of new potentially toxic/pathogenic ones. It was praised as a breakthrough revolutionizing medical research. But surely OP already heard of this. I guess that's why he's asking specifically about OpenAI and not AI in general. No, OpenAI hasn't yet announced a project in this area.
I remember that. Not sure what its applications are but Ive always saw promiae in potential medicines of aome sort, reducing the need for testing and trial and errors would be monumental for to-market Rx.
shit post
This was a great exaggeration
When actually tested almost none of the predicted proteins were correct
Why are you all such idiots?
>almost none of the predicted proteins were correct
Source? I canna find anything on Alphafold or Alphafold 2 predicting proteins incorrectly.
>Source? I canna find anything on Alphafold or Alphafold 2 predicting proteins incorrectly.
NTA, depends on what you mean by prediction. It's complicated. Its real success was with https://en.wikipedia.org/wiki/CASP which is just trying to identify protein 3D structure from its amino acid sequences.
Where these things fail catastrophically is attempting to predict previously unknown protein structures from amino acid sequences. To me that just screams "overtfit model to beat the test without generalizable application beyond the test". It doesn't really get us closer to the real holy grail https://en.wikipedia.org/wiki/De_novo_protein_structure_prediction
Also with respect to whole system modeling of complex biological organisms and therefore prediction of those results, uh, we ain't very close to doing that. Hence why the GPT thing is funny. Though plenty of probability assorting algorithms and tools exist for doctor assistance in diagnostics, they absolutely are not to take the place of actual testing or having a competent doctor with a brain to know when the suggestions are full of shit.
See also https://en.wikipedia.org/wiki/AlphaFold#Limitations
There's definitely AI being used in biology, I'm not really sure of the extent of it because I'm not in that field. But there's lots of papers mentioning AI
https://www.biorxiv.org/search/Ai
If seen a few papers where they were using data science too, like machine learning etc so I'm pretty sure it's all being used but maybe there's no major breakthroughs yet I'm not sure
>Why isn't openAI used in biology/medecine?
Because it makes shit up.
the FDA won't approve its wild spread use because it will make 80% of medical doctors go jobless. they will lobby against AI as long as they can.
You can't simulate something if you don't even know how it works in the first place.
Why do so many people think that computers, especially when you vaguely toss around the concept of "AI," are magic?
Thank god she's out of the white genepool.
No wonder she couldn't find a date. Look at the colour of her kid!
>Why isn't openAI used in biology/medecine?
because it's a moronic llm that predicts response based on its data and not an agi. it can't know , infer or discover anything on its own, it has to have that information in its dataset.
ML is already widely used in biology.
>we are kings and queens
why didn't she take her baby daddy to prom?
does this mean short niglets? How do I profit from this?
Lots of large learning systems and other sorts of "AI's" have been and are being used in biology and medicine.
Just not OpenAI, because it's a chat bot not suited to the task.
because it's like four months old
Why aren’t cars used in biology/medecine?
I believe the potential would be much greater there, instead of driving on roads, they would try to figure out how human cells and body work by simulating embryonic cells.
not only it would help cure most diseases and potentially aging it would also remove the need for animal testing.