If the inference is profitable and training new models is actually an endless arms race that's actually the best outcome for nvidia specifically.

Here is a charitable perspective on what's happening:

- Nvidia has too much cash because of massive profits and has nowhere to reinvest them internally.

- Nvidia instead invests in other companies that use their gpus by providing them deals that must be spent on nvidia products.

- This accelerates the growth of these companies, drives further lock in to nvidia's platform, and gives nvidia an equity stake in these companies.

- Since growth for these companies is accelerated, future revenue will be brought forward for nvidia and since these investments must be spent on nvidia gpus it drives further lock in to their platform.

- Nvidia also benefits from growth due to the equity they own.

This is all dependent on token economics being or becoming profitable. Everything seems to indicate that once the models are trained, they are extremely profitable and that training is the big money drain. If these models become massively profitable (or at least break even) then I don't see how this doesn't benefit Nvidia massively.

The best bots are from cheating rings where the bots are colluding, not from bots that play poker perfectly (which don't exist in any real sense for full ring poker).

What's true is that people aren't making as obvious mistakes, especially preflop, so you can't make hundreds of thousands just by knowing that Ace King is a good hand that you can go all in with. Anyone can find preflop poker charts and fix part of their preflop game.

Having a poker solver isn't enough. Let's say you play tournament poker, just having a basic understanding of concepts like ICM give you a massive edge. Let's say you take it a step further and understand concepts like "future game" and actually study them using tools, you're edge has expanded further.

There are a bunch of charts out there that tell you what hands to go all in with if you have 15bbs or fewer. None of those charts take into account ICM. Also how do you adjust the charts if your opponents are calling with too many hands? How do you adjust them if they call with too few hands?

Let's say we are just talking about cash game poker, it's not enough to have a solver, you need to understand how to actually study with the solver. People try to use them like a cheat sheet that tells you what to do, not understanding that a slight change to the inputs of the solver can drastically change the output. The purpose of a solver is to understand how different ranges interact at different stack depths on different boards.

ie: Playing 100bbs deep, on a KK3 flop with a flush draw, what hands should i check or bet as the preflop raiser? What happens if that 3 is a 7? What if it's a J? What if it's 33K instead of KK3? What if I'm 200bbs deep instead of 100? What if the opponent calls too much? What if they call too little?

I'm literally winning money playing online today playing 400nl (200nl with a straddle, in the US).

Please explain to me how you think these bots work? Do you think they are literally hooked into solvers and solving these hands in real time? If you actually understood poker you'd understand that the winrate from GTO is not good enough to make real money playing poker without a massive sample size, the game is all about exploiting players when they deviate from GTO. Explain to me how you program your poker bot to know intuitively that a player has too many bluff combinations when a flush arrives on the turn after they check back on the flop therefor you should call wider than standard? There are a billion little unique situations where people don't bluff enough, bluff too much, call too much or call too little and that is where the winrate from poker comes from.

This is the difference between having a 3 bb / 100 winrate and a 10-15 bb / 100 winrate. Maybe there are a bunch of shitty poker bots winning at 1 bb / 100 but if they are winning it's because some players suck really really bad, not because they are playing perfect poker.