“There are two ways work can compound. It can compound directly, in the sense that doing well in one cycle causes you to do better in the next. That happens for example when you’re building infrastructure, or growing an audience or brand. Or work can compound by teaching you, since learning compounds. This second case is an interesting one because you may feel you’re doing badly as it’s happening. You may be failing to achieve your immediate goal. But if you’re learning a lot, then you’re getting exponential growth nonetheless.
If you’re not learning, you’re probably not on a path that leads to superlinear returns.”
“Something that has changed a lot since 20 or 30 years ago is that a lot of people are making records by themselves. I feel like the implications of that are not often thought about. I don’t know that the market has ever really shown, in the streaming era, that there’s an enormous amount of demand for that type of expression. […] The whole idea of the recording artist as somebody who sits in a room by themselves and the rest of us are supposed to see that this type of expression is interesting and valuable and important by default–I don’t know if in the streaming era people are still seeing it that way…We have to ask ourselves: Does this stuff have value?”
“Search engine summaries at least offer users links and citations and a way to link back to Wikipedia. The responses from large language models can resemble an information smoothie that goes down easy but contains mysterious ingredients […] The ability to generate an answer has fundamentally shifted…in a ChatGBT answer there is literally no citation and no grounding in the literature as to where that information came from.”
“When we seek knowledge, we don’t just blindly consume ever-larger data sets. Instead, we have things we want to know. Taylor Beck, a neuroscientist turned teacher, once pointed out to me that A.I. might be the only context in which you find truly unmotivated learning: the machine just ingests a mass of undifferentiated text, none of which it cares about. Natural intelligence, by contrast, is almost always accompanied by some want, or at least a goal—whether it’s a toddler in search of joy or an E. coli bacterium that, because it ‘wants’ to eat, performs a sophisticated computation measuring the chemical gradients in its environment. In this view of intelligence, drive is primary […] A major leap in A.I. may come when L.L.M.s start seeming curious, or bored.”
brettworks 
Leave a comment