1,749 mentions · 790 chunks · 131 episodes
If you come up with a definition of a task LLMs can't do... can humans? When you understand a machine or animal better, it causes you to reflect on your own human skills. What makes us human? "LLMs...
... over time, entropy eroding it. Humans absorb knowledge into language. And then LLMs can come along and slurp up that knowledge. LLMs and neural nets are extremely rudimentary processes, just in an extremely large scale system soaking...
...hat a weird distillation, what a compressed pipe of information. That's the way LLMs see the world!
...ighting the transition because we want the control that comes from building it. LLMs get their abilities organically, not via engineering. To unleash the power of LLMs, we have to move to a gardening mindset, not a builder mindset.
LLMs are like a magical photo copier. They can do a surprisingly good copy of things they've never seen before. But each thing they generate is slightly d...
LLMs have a "cool teacher" voice and say the most bland things. (Riffing off of Molly White's excellent https://www.citationneeded.news/ai-isnt-useless/) ...
LLMs are inherently bland. If you ask ChatGPT to ask you an interesting question it'll say something super generic like: "What movie or book do you think ...
LLMs don't reason, they intuit. With enough scale, this can do an extremely convincing facsimile of reasoning. LLMs appear to be possibly incapable of ori...
A pattern: use LLMs for a rough and ready version. A quick and dirty, good enough answer on demand. The more that procedure is depended on, the more you factor out commo...
...e whole system should be possible for humans to do, even if it's a lot of work. LLMs then provide some good-enough starter ability. The LLMs are the floor and the height of that floor is how good the LLMs are. The higher the floor, th...
The app model has no wiggle room. Users being able to use LLMs to jury rig solutions doesn't change anything inside the app model. But outside the app model, being able to jury rig solutions is obviously useful.
LLMs are magical duct tape that are made out of crystallized society-scale intuition.
...ally lower-cost-than-before input. The low-cost input for this next paradigm is LLMs. LLMs look expensive compared to normal compute. But they look radically cheap compared to generic human mental labor.
LLMs mean that anything with an API can now be controlled in plain english. There are a ton of amazing open source tools and frameworks that previously we...
LLMs in the computation loop can create more resilience. Last week I riffed on the idea that systems with humans embedded are more resilient. Systems with...
LLMs allow you to talk to the crystallized intuition of society. LLMs: a deterministic system that outputs vibes LLMs are the apotheosis of the algorithm....
Today on Twitter, people doing cool things with LLMs share screenshots, not running links. That's because there's no good way to distribute it. If you link to a hosted demo with your own API key, you'll...
...or AI. The consumer app model requires services supportable by advertising. But LLMs are too expensive to be supported by advertising. So you need a subscription… but how many subscriptions will a user pay for? The one-size-fits-all, ...
Building a bit more on the LLMs as trained circus bears analogy from last week. The trajectory we're on as an industry is not better-trained circus bears, but more of them on the lo...
...eed to figure out a way to work with it productively given that it's untrusted. LLMs are gullible and squishy, and highly susceptible to their (perhaps hidden to you) inputs. You must treat LLMs as an untrusted component in your syste...