What you want is a distribution-aware contextual binary search. With whatever information you have (appearance, personality, vocabulary, etc), you can come up with a probability distribution in the space of possible ages and start your guess with the value at the 50th percentile. Then depending on whether the true age is higher or lower, your next guess will be either the 25th or 75th percentile. Rinse and repeat.

In reality, the way most people intuitively do agree guessing is already an approximation of this procedure.

How do you start on finding an alternative without first acknowledging the flaws of the current system?

You can also tell me that someone out there won the lottery this week and have it be true. It’s not the same as seeing this person’s live reaction to learning about it. It wouldn’t be the same if you watched that person act out the scene exactly as it happened. AI generated is so much further removed from all that.

The info provided is that there exists another happy dog out there doing happy dog things and I briefly connected with it, which made me happy. This information would be incorrect if it was AI generated.

So what you’re saying is that Andrew Yang should run again.

Part of the reason Amazon works well is because they sell high volumes of each product, allowing them to distribute products ahead of time across warehouses to match expected demand. You can’t do that if you only have exactly one of each item.

In a socialist country, I would guess that there’s much less incentive to pump out slop. So if you make videos with AI, it’s more likely that you’ve actually put some thought into it and are making something of actual value to someone.

Plus, statistics make up the basis of pretty much all of our science. If you dig into the foundations of stats, you’ll find that it’s basically just formalizing our feelings. It just happens to be formalized in a way that appears to reflect reality accurately enough to be useful.

Did Xi actually take offense to it? I thought it was just others being overly heavy-handed in their censorship, thus Streisanding the whole thing.

I’ve always just done lots of water and waited it out. I see medication at the pharmacies labeled “cold medication”, but I never looked into what they do.

Meaning that it’s just marketed as cold medication without doing anything specifically for colds?

For each tab, I find the project(s) associated with it, find my notes for that project, save the URL for that page in the appropriate place in my notes, then close the tab.

If it’s something that isn’t for a specific project (e.g. reading something because it looks interesting), then I just close it. It’s not important. There’s plenty of entertainment to be found without those.

The difficulty in training an AI is dependent on data availability. So this is just a question of choosing a language that has the least amount of writing. You can trivially choose any language that doesn’t have a writing system at all and invent a writing system for it. But then you’d also run into the problem of learning the language yourself.

Maybe it’s “single” as in “one unit of mother”

The main difficulty is in how many hyperparameters are involved in training an RL agent, high sensitivity of RL algorithms to those hyperparameters, and not having a good understanding of how to select them based on the properties of your task. This problem is exacerbated by the high sample complexity of RL. If something doesn’t work out, you don’t know if it’s because you chose the wrong set of hyperparameters or if you just haven’t trained for long enough.

I don’t know much about game design, but I do know that it’s a much more mature field than RL, so surely they have better tools than guessing and praying.

It is expensive, but it does work. We’ve already seen things work to a limited extent on StarCraft 2, Dota, and Gran Turismo, and those are all multiplayer games. The article seems to be talking about single player games, which simplified things a lot.

Game playing is not LLM. They’re game-specific reinforcement learning models. It’s not easy, but definitely doable with existing tech. Sony’s GT Sophy is a good demonstration on what they’re capable of.

I don’t know if you can describe it as “can’t be arsed” when their proposed solution is so much harder to implement.

It’s not an analogy. It’s a counterexample. One that is irrelevant because I appear to have misunderstood your argument, but you’re not clarifying, so I have nothing new to add here.