88 chunks · 61 episodes
...if you have a very long runway. The risk is that the longer you haven't touched ground truth the more likely you are to never be able to touch down again, at a compounding rate. You get more and more out of touch and then when you try to touc...
...arring match because you're right rather than because you're powerful helps you ground truth your beliefs. His manager saw the value in that for truth-seeking.
The feedback loop on model improvement is "how long does a ground truth check take" Do you need to talk to a human? Do you need to run a physical experiment with expensive inputs? If it can be done entirely at computer sp...
...ost of the input is junk. In 90 iterations, all of it is. LLMs decohere without ground truthing with real world results. That can be done automatically with things like React components. But anything that is the least bit complex has to be pu...
...iver. You can still get some emergent signals, but they are warped away from a "ground truth" of quality and can get warped by political dynamics. If there's no cost to them doing it, then the signal is superficial and performative. Meaning c...
... that doesn't have to fight to survive succumbs to kayfabe. The larger and more ground truthed the organization, the longer this process takes.
... time with LLMs than people, you could decohere from society and truth, with no ground truthing of "no, I don't think that makes sense…"
...ative even stronger. The only way to counteract this is to have the org have to ground truth constantly. An org that is so successful that they can operate poorly for years without dying but that fires people easily will have the worst of bot...
... motivated views of the same scenario, which makes it easier to triangulate the ground truth.. Perhaps Granola style transcripts could help ground truth? But only if everyone's actions didn't change given they were being recorded. Do we try t...
...ledge for what you need. Too much fidelity, not enough doing. The doing is what ground truths and helps figure out where your mental model is incorrect. It's the same asymptotic curve of perfection that is a bad idea in building products. Ove...
... unintentionally). Who should be the singular arbiters of the infinite, perfect ground truth? There is no answer. "Balancing" can itself also be bias. A point someone made this week: "If one candidate's policy was to encourage infants to eat ...
...limit. Companies are a swarm of employees making interdependent decisions whose ground truth is the business model. Ultimately the ground truth about what a company cares about is the business model. Everything else is just platitudes.
In a post truth world ground truth doesn't matter in any given interaction. But of course the ground truth never goes away; it will matter at some point. The first person who points ou...
...olitical players. The only thing that contains the expansion of politics is the ground truthed success of the company in the wild. Imagine if you had a swarm of LLM-backed agents in a company. They could conceivably absorb infinite energy, al...
LLMs will get good at everything that can be quickly ground truthed by computers. But that doesn't mean they'll do a good job at things that can't. Most meaningful things to humans can't be ground truthed entirely i...
...order propagates? You need a percolating sort; a thing that uses some energy to ground truth and rank so the good things naturally survive to the end.
...evaporates. But every so often one that is better is found; it has a consistent ground truthed bias that makes it stand strong amidst the noise. This is conserved, naturally. The thing that tried it keeps trying it while it works. Other entit...
... hypothesis. Then expose that hypothesis to disconfirming evidence, for example ground truthing it. Guess and check. The disconfirming evidence gives you the information necessary to update your hypothesis to make it more accurate. If you don...
User feedback should be used as disconfirming evidence. It helps test and ground truth your hypothesis. it doesn't tell you what to think. You have to have your own hypothesis. What your early adopters (or users via UXR) ask you to do i...
...herence startup can't have any kayfabe. It needs to be aggressively, constantly ground truthed in order to survive. It's only post coherence organizations that can (temporarily) survive kayfabe.