On one hand, don't be the tech guy asking "tough questions" in business meeting about advanced tech, because you are the a-hole (have been that guy, have been on the other side as well)
On other hand too much incentive to fool CEO who doesn't know overfit
I'm all for demos. This shows the upper bound, and lets you dream a little. We all need that.
But ML isn't like building a car. If you build the demo car, clearly more cars can be built (at some cost). The steps taken toward creating a compelling ML/AI demo can be completely divorced from making it work in practice.
I wish I had good examples... but this is really just about hard-coding, and overfitting.
Think having a fake/choreographed fight vs real boxing match. Ok... maybe bad example 😂 @ana_analytics_
There's skill to making a really good fake AI. But it's different from the skill in making a decent real one. The guys who are wizards on And-One tour, by and large had no chance to make the NBA. Even though they do things the NBA guys can't...
I get that some in USG don't like Russia. They just want Russia to lose, at everything. I get that impulse. Not sure I see a deeper justification here.
As much as it hurts me to say -- as an applied deep learning guy for years, and still a massive fan of the space -- in this climate, "AI for X" startups might not do very well. Even if you raise, you may not have the room to figure it out.
I still think AI/ML/deep learning will be solving a lot more practical problems. This will be a long-term trend over the next decade+. Not at all a bad space to get into. But you've got to have a long term approach.
I think ML infra and tools are good. But some of those companies will also struggle, as their customers struggle, and BigCo will be more reluctant to throw people and $ at the space without a clear goal in mind.
Give yourself respect if you can understand the complex biology of how they control individual neuron cells. Explained really well, but it's intricate. Gonna listen to it for third time now. It's well worth it. 🧬
Amazing where biology has gone in the past 10-20 years. Growing up in a physics household, it was not that hard sciences looked down on biology (they did not) it's just biology wasn't very precise. We were amazed what they could do with such un-exact tools back then.
Now the level of measurement and control is kind of unbelievable. You can hear that in @PeterAttiaMD voice, and he know the work.
Same as every time @razibkhan describes what's possible with ancient DNA.
The kind of science we could only dream of as kids.
O1 has always been a flexible but mysterious program. Our family got it -- but easy since my father was physics professor at Tier 1 university.
Had Russian friends who got it through chess -- GM/IM pretty easy -- you have international titles.
Seems it was always focused on credentials. That isn't changing but I guess now University cred (PhD) is enough, as opposed to needing some central body to certify you (like an academy of science, international conference, FIDE, etc)