Daniel George
I care about improving the rate at which we do science. The bottleneck I'm currently interested in is tacit knowledge capture.
In the past, I worked on creating tools that enable us to have thoughts we wouldn't have had otherwise. My serious context of use has been in biology.
Previously founded Speck (YC W24) where I worked on realtime AI assistants and was an engineer at Sphinx Bio (acquired by Benchling)
Selected Projects
Biomemex
Prototype for retrospective analyses on wetlab recordings with AI glasses in a similar vein to Vannevar Bush's Memex.
We'll have AGI before we have humanoids so they'll need interfaces to work with us to interact with the world. A form factor like this one could be a possibility for that. The POV data is also a hedge against humanoids since you can use it to train them as well.
Ambient AI Doctor
Turns every piece of data from your life (ex. food pics, workout info, voice notes) into a structured format so that an AI doctor can make correlations and analyses.
The bottleneck to having AIs optimize our lives is having the data to reason on. Lowering the activation energy of ingesting tidy data is the most important problem to solve here.
Spheroid Analysis
Spheroids are 3D cultures that mimic tissues and micro tumors better than the 2D cultures we see in Petri dishes.
When extending biological circuits from 2D to the more realistic 3D, the morphology of the cells could affect things like protein production. Protein synthesis is an inherently stochastic process, so when creating biological circuits quantifying a measure of noise in the system is quite useful. This is extending the 2D analysis to spheroids.
Perceptual Filter
There have been many demos of LLM filters for feeds but I believe this was one of the first! It was impractical due to how slow and imprecise the models were, but was directionally correct and more important than ever.
You define what's important to you and it reasons whether a tweet in your feed should show up.