Hearted Youtube comments on Dr Ben Miles (@DrBenMiles) channel.
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
4
-
Great video! Deep tech is a fun word! I worry students are discouraged from pursuing deep tech as a result of deadlines, risking the entirety of your student loans, education and career on a moonshot paper idea seems stupid. Especially when most of the potential rewards are so far removed from the researcher.
It reminds me of a video on neural nets that I watched, an AI in mario kart was given a -100 penalty for falling off the edge of the track, but not enough of a reward for moving forward, so the AI learned their best bet was to just sit there, or to move extremely slowly, because it was too scared to take any risks, as a result of the penalties far outweighing the potential rewards. Mario learned to drive forward after the penalties where reduced or incentives increased. (Source "MariFlow" & "MariQ")
Your distinction between shallow and deep tech was really spot on, I personally find the difference to be whether or not you can build your invention with parts that are already mass produced, and simply assemble them together, or if new recipes and factories must be designed and built to mass manufacture the components necessary to build your device.
It's easy to buy a bunch of components and assemble them, it's a lot harder to design and build the components to then assemble into a single object. On a software level, this means any program that simply interfaces between preexisting libraries would be shallow, whereas the requirement of brand new untested algorithms would be deep.
The best way to do deep tech in my experience is incrementally. Lets say I need IPMCs for my MMRs, it's in my best interest to mass produce and sell IPMCs to other companies, then start incorporating them into my robots. Don't try to do too many new things at once, go one step at a time and make a profit along the way.
If anyone of your startups ever mass manufacture IPMCs give me a call, I need a lot, so does the rest of the world.
4
-
4
-
4
-
4
-
3
-
3
-
3
-
3
-
3
-
3
-
3
-
3
-
3