AlphaFold uses a combination of machine learning and physics simulations to predict the shape of a protein given its chemical formula.
Gemini is a Large Language Model which uses machine learning to predict text completion.
They both use “machine learning” but the machine learning models are wildly different. Machine learning is a technique that’s basically a fancy curve fitting approach. The curve AlphaFold is trained to mimic maps chemical formulas to shapes, while the curve Gemini mimics maps a text input to probabilities for predictions of the next letter(s).
Funny, one of the first arguments I had about AI was with someone who insisted it would be incredibly helpful at predicting protein folds.
That person was right: https://medicine.iu.edu/blogs/research-updates/the-protein-folding-problem-the-day-ai-unlocked-a-secret-of-life
So this may be why Google is closing AlphaFold; AI is already doing it.
AlphaFold is Google’s AI approach to protein folding…
I meant that I thought they were merging the two if Gemini could simply do what AlphaFold had been doing.
Gemini is a completely different project.
The term “AI” is rather overloaded these days.
AlphaFold uses a combination of machine learning and physics simulations to predict the shape of a protein given its chemical formula.
Gemini is a Large Language Model which uses machine learning to predict text completion.
They both use “machine learning” but the machine learning models are wildly different. Machine learning is a technique that’s basically a fancy curve fitting approach. The curve AlphaFold is trained to mimic maps chemical formulas to shapes, while the curve Gemini mimics maps a text input to probabilities for predictions of the next letter(s).