It’s not an llm. Thank you Jesus. :jesus
Well yeah, it’s China. A US-backed product would be closed-source, involve throwing money into a flaming trash barrel with Nvidia’s logo on it, and would only be used for finding new formulations of boner pills and hair regrowth serums.
I missed this tech being open source. Is that in the article?
It’s mentioned in the abstract of the linked paper, but that blurb sounds more like they’re talking about the data set that they used for testing it. Based on past Chinese AI projects (e.g., DeepSeek), I guess I just assumed that the DrugCLIP framework would be, but at least the code used for generating the results in the paper is available on GitHub: https://github.com/bowen-gao/DrugCLIP
I’m once again reminding that synthesis and testing on the cells->tissues->mice->monkeys/piggies->humans takes the bulk of time and money in drug development, the screening has long been kinda not a barrier
That’s not to say it’s not good that they did it, it’s that this will shave off maybe 3 months from 5-7 year process, and probably produce myriad troll patents somewhere
Hmm, that’s an interesting approach
