Question: creating babyimages from genomic data
0
Neo • 0 wrote:
is anyone working on an a.i. build to assemble assumptions for future babies? like is it possible to use clustal analysis to reconstruct future baby faces from training data only? my idea was that you take the genomic data of both parents and the pictures of their babies to train an ai to build those pictures from scratch
The human genome is more than 3b nucleotides, in theory it is possible only if we know all nucleotides that determine the face shape! It is even harder to obtain genome sequencing data and the matching faces for children and their parents.
Similar AI challenges with available data is predicting nucleotides that cause genetic diseases in children such as mendelian disorders.
Maybe that's something for google? As they already beat everyone on protein structure prediction. The idea would be certainly intriguing and a bit concerning also. I think at the moment, nobody can say for sure if the prediction of faces - of babies and of course also adults - from DNA will give very accurate results in the future, but it might very well be. Might need to take some epigenetics data into account as well.
Just out of curiosity about the protein structure prediction tool by Google (AlphaFold ): has anyone used it yet? It looks very glossy and promising (like all Google products do), but has anyone applied this in their respective research, or tried bench-marking it with available tools (Modeller, i-TASSER, etc)?
No, I don't think they have released it. They just came, showed off in the contest, and didn't deliver :)
i will be building a server explcitly for this problem in the next weeks. Will keep you updated in this thread. if anyone wants to join he is pleasently invited.
Make those baby faces!