robotelvis

Cars are a long way from achieving Artificial General Horsiness (AGH). There are many things a horse can do vastly better than a car, such as jumping over fences, navigating by itself, swimming through water, fueling themselves from naturally occurring vegetation, and providing social companionship.

And yet cars have managed to replace horses for pretty much all tasks that people previously used horses for. This is because it turned out that, while cars can’t do everything horses can do, their speed and cost advantages were big enough that it was worth adapting the rest of the world (e.g. building roads and gas stations) to compensate for their limitations.

There may be lessons here for how we should expect AI to replace human labor.

That's already the de-facto situation on iPhones a lot of the time - you can set it to only ring when called by a contact.

The downside is that I've missed important calls with that setting on. I'd love the option to opt into only allowing calls from some kind of verified white-list.

Or we could just have an FCC that does it's job.

GPT-4 was created by humans, but is too complex for humans to understand. This puts it in a broader class of valuable-but-incomprehensible things that includes the world economy, the source code for large software products, and life itself.

GPT-4 was created using a rapid iterative process where vast numbers of GPU cores made tiny iterative improvements to a model of human knowledge. Indeed Matt Ridley has argued that a similar evolutionary process is responsible for the creation of pretty much everything that matters.

But this kind of productive evolution doesn't just happen automatically by itself. You can't just pour out a pile of DNA, or people, or companies and expect something good to magically evolve. For productive evolution to happen you need the right structures. Those structures ensure that iteration is fast, that each step has low risk of causing harm, and that iteration progresses in a good direction.

I wrote a blog post about how those structures work, drawing on examples from machine learning, metric driven development, online experimentation, programming tools, large scale social behavior, and online speech.

I’ve done ranking-adjacent work at three companies (Google, Quora, and Facebook), so when OpenAI released the ChatGPT API a few days ago, my natural first instinct was to ask myself “I wonder if this would be good at ranking social media feeds?”.

Based on my experiments so far, it looks like the answer is “yes”. Indeed GPT makes content-based feed ranking so easy that it has the potential to completely change the way we think about ranking - potentially shifting ranking power from platforms to individual users and groups.

I was actually the "author" of one of the pieces of code that SCO claimed had been stolen from them - SCO's implementation of the Berkeley Packet Filter.

SCO had a collaboration with IBM called Project Monterey, in which SCO and IBM were merging their operating systems in order to be better positioned for IA64 (which at that point people still thought would be good).

One minor detail of the Monterey deal was that SCO got the right to port IBM's superior networking stack into SCO's operating systems. One part of IBM's networking stack was their implementation of the Berkeley Packet Filter, and that was one of the parts I ported over (I'm pretty sure the file still had a BSD copyright notice on it).

That was back in the days when SCO was a cool UNIX vendor that thought of itself as an extension of UC Santa Cruz. Back then employees couldn't really imagine it turning evil, and it made me wary of what my subsequent employers (Google and Facebook) could turn into.