How about a co-op of scientists pooling money to rent the kinds of AI chips the big players are starting to put out? I’m sure there are holes in that argument too; but c’mon, this is hacker news. If we don’t try to overcome the hard challenges then what’s the point of this community? ;)
> How about a co-op of scientists pooling money to rent the kinds of AI chips the big players are starting to put out?
What is really expensive and complicated for now is “wet” data collections: it is a finicky process stuffed to the brim with very expensive machinery, consumables, lab space and manual workers. Computing power is not really a limitation nowadays, you can comparably get a whole lot of bangs for your bucks.
Concerning the sharing part, equipment pooling is definitely already happening, at least at the local scale. And what is getting more and more developed is the “platform” concept, where some labs/teams slowly transition from doing research to producing data according to a standardized protocol (genome sequencing, genotyping, ChIP-seq, etc.) for other teams doing the research.
> I noticed companies moving into the space, like 23andme
23andme is not doing sequencing, they're doing genotyping. And they are doing so using relatively old and well-established method; so I highly doubt they will be a force of innovation in the sector for the coming times.