today's models of artificial intelligence like GPT 2 and GPT 3 can write entire articles, books, and poems.

Why not make such a generator for proteins?

The generator creates a story from the letters of the protein, where the laboratory calculates and synthesizes it. I think there will be interesting results.

for the training set I think it is worth using the PDB database

How do you feel about the idea of AI GPT 2-3 ( text AMINO ) creating new proteins?

>there are more than 100 thousand genes in the human genome.

Baker's lab could take 100,000 genes from human dna and train AI to work with dna.

Such an AI could write future dna, and future protein sequences that will occur in the future