Post

I do firmly believe that refusing to outsource your basic cognitive functions to LLMs right now - as so many are stumbling over themselves to do right now, like hogs trotting to a pile of slop that’s been set up inside of a butcher’s delivery truck - will very much pay off in a few years.

And by this I don’t even mean “refusing to use LLMs for anything at all” - they do have some use cases - I mean “refusing to outsource huge swaths of your day to day brain functions to the lying machine.”

I've so far seen 2 solid use cases of AI in my line of work: generating a first impression of spectral flow cytometry panels that are unmixable and halfway decent first runs at scaling the data derived from that. Both directly downstream of significant amounts of expert curated data. And that's it.