@TheStalwart randomness (for variety, and to reflect that the probabilities are estimates) and various search strategies (to consider not just next token, but phrases and more). That’s the sampler, which isn’t really part of the LLM, but in practice is essential.
@TheStalwart The output of the model isn’t a token, but a probability for every token. Imagine how choosing the most likely next word in all circumstances would turn out: no surprise or interest, always the same output. So that part is deterministic, but not very useful. So we layer on…
@nateberkopec Lol... I've always liked BDD, maybe because I do contract work and my work needs to survive a relay of future contractors. We need these high-level, common sense (obvious?) guardrails for the verbose, naive code bots.
@mattracquet Mixing in drop shots can force an opponent to position differently and think about different types of shots. Does that show up in the data at all?
@EamonnDwyer@robinhanson Makes sense, hadn’t thought of radon. I’m running that experiment on myself here in Tennessee, ha! Was wondering about high elevation as well. There’s a positive association with longevity, and I don’t think the mechanism is known (theories include lifestyle and hypoxia).
When I was on the ISS for my nearly year long mission, there was a telomere experiment comparing my telomeres to my earth baseline and my twin brother as a control. Hypothesis was they would get damaged and worse due to the environment. Turns out they got better. Initially NASA thought maybe it was due to exercise and diet. After I returned we learned JAXA had a telomere experiment on some small worms the same time I was there. Their telomeres got better too. Never saw the worms doing any exercise. After further study determined it was the radiation.
@robinhanson Given this and Scott Kelly's story, has anyone analyzed frequent fliers. I've heard for years that there's an increased dose of radiation at high altitude. But maybe it's good? Someone ought to compare pilots and flight attendants to their siblings.
@GaetenD I don’t know what the specific market is, but as a dad of an 8 and 6 year old I can say that Golden plays go to infinity because children don’t get tired of songs, no matter how many times you play them.
@_xjdr This naturally incentivizes testing, especially at the system/integration level (cool), incremental changes and regular merging (cool), and microservices (yikes).
@rohindhar If one had the money and will to play the long, long game, what would that look like with one of these? I see them, and I wonder if anyone is intentionally letting them go vacant. Are there alternatives, or are these properties doomed to perpetual decline?