3 ms·
Disagree about bootstrapping AI tech. People who can actually implement, or build neural nets from scratch in a scalable system are in massive demand and are no
by AndrewKemendo 10y ago
Disagree about bootstrapping AI tech. People who can actually implement, or build neural nets from scratch in a scalable system are in massive demand and are not cheap - from a time or money perspective.
To build competitive nets you need a way to acquire data (not cheap), label it (also not cheap), and train nets with people who know how to implement them (very much not cheap).
That's all before you have a product which is generating revenue to "boostrap" from. Even Geohot needed 3M from Andreesen to get comma.ai off the ground to gather and train data from a free app and try and turn it into a product (which isn't out yet).
I'm talking about $10M here, not $10B. You max out Angel money around $1-2M which would get you maybe a year with some MS/PhD Machine Learning Engineers. I know this from first hand experience. And yes, you really do need people with that level of background. I've never met a self trained ML expert who can rapidly build ANNs and gather the data efficiently (though I am sure a few exist).
- Retric 10y agoYou can learn how to leverage this stuff without a PHD, and if you have a PHD you can be one of the guys doing the bootstraping. Also, NN are also only one kind of AI company. Squirrel-Tracking Water Cannon guy is one example where someone use basically off the shelf components to hack something together. Finding the right sized niche is hard, but that's always been the core problem. How about showing what a new couch might look like in a room from a cellphone video. There are already several things that do most of the heavy lifting can you glue them together into an MVP? Is there already a product that does this? Can you get referrals from IKEA? etc.