Kevin Lacker

Posts

Sorted by New

Comments

AGI Predictions

I suspect this question is misworded:

Will there be a 4 year interval in which world GDP growth doubles before the first 1 year interval in which world GDP growth doubles?

Do you mean in which world GDP doubles? World GDP growth doubles when it goes from, say, 0.5% yearly growth to 1% yearly growth.

Personally, I suspect world GDP is most likely to next double in a period after a severe war or depression, so you might want to rephrase to avoid that scenario if that isn't what you're thinking about.

Persuasion Tools: AI takeover without AGI or agency?

I believe there already is a powerful AI persuasion tool: the Facebook algorithm. It is attempting to persuade you to stay on Facebook, to keep reading Facebook so that their "engagement" metrics are optimized. Indeed, many of the world's top AI researchers are employed in building this tool. So far it is focused much more on ranking than on text generation, but if AI text generation improves to the extent that it's interesting for humans to read, I would expect Facebook to incorporate that into newsfeed. AI-generated "news summaries" might be one area that this happens first.

Hiding Complexity

I don't think the metaphor about writing code works. You say, "Imagine a company has to solve a problem that takes about 900,000 lines of code." But in practice, a company never possesses that information. They know what problem they have to solve, but not how many lines of code it will take. Certainly not when it's on the order of a million lines.

For example, let's say you're working for a pizza chain that already does delivery, and you want to launch a mobile app to let people order your food. You can decompose that into parts pretty reasonably - you need an iOS app, you need an Android app, and you need an API into your existing order management system that the mobile apps can call. But how are you going to know how many lines of code those subproblems are? It probably isn't helpful to think about it in that way.

The factoring into subproblems also doesn't quite make sense in this example: "Your team still has to implement 300k lines of code, but regardless of how difficult this is, it's only marginally harder than implementing a project that consists entirely of 300k lines." In this case, if you entirely ignore the work done by other teams, the Android app will actually get harder, because you can't just copy over the design work that's already been done by the iOS team. I feel like all the pros and cons of breaking a problem into smaller parts are lost by this high-level way of looking at it.

My null hypothesis about this area of "factored cognition" would be that useful mechanisms of factoring a problem into multiple smaller problems are common, but they are entirely dependent on the specific nature of the problem you are solving.

How do you evaluate whether a $500 donation to a project that you know well is a good idea?

I would not spend $500 on such an event because an event held by my local rationality community doesn't seem very important to me. You may have a different opinion about your $500 and your local rationality community and that's fine.

Should we postpone AGI until we reach safety?

We could in theory try to kill all AI researchers (or just go whole hog and try to kill all software engineers, better safe than sorry /s).

I think this is a good way of putting it. Many people in the debate refer to "regulation". But in practice, regulation is not very effective for weaponry. If you look at how the international community handles dangerous weapons like nuclear weapons, there are many cases of assassinations, bombing, and war in order to prevent the spread of nuclear weapons. This is what it would look like if the world was convinced that AI research was an existential threat - a world where work on AI happens in secret, in private military programs, with governments making the decisions, and participants are risking their lives. Probably the US and China would race to be the first one to achieve AGI dominance, gambling that they would be able to control the software they produced.

Libertarianism, Neoliberalism and Medicare for All?

If the government is going to mandate something, it should also pay for it.

This isn't really how government mandates work. The government mandates that you wear seat belts in cars, but it doesn't pay for seat belts. The government mandates that all companies going public follow the SEC regulations on reporting, but it doesn't pay for that reporting to happen. The government mandates that restaurants regularly clean up the floor, but it doesn't pay for janitors. The government mandates that you wear clothes in public, but it doesn't buy you clothes. Etc, etc.

So I think your intuition is simple, but it largely does not map to reality.

Inner Alignment in Salt-Starved Rats

Now, for the rats, there’s an evolutionarily-adaptive goal of "when in a salt-deprived state, try to eat salt". The genome is “trying” to install that goal in the rat’s brain. And apparently, it worked! That goal was installed! And remarkably, that goal was installed even before that situation was ever encountered!

I don't think this is remarkable. Plenty of human activities work this way, where some goal has been encoded through evolution. For example, heterosexual teenage boys often find teenage girls to be attractive and want to get them naked, even before they have ever managed to do it successfully, without a true conscious understanding of their eventual goals. Or babies know to seek out nipple-shaped objects, before they have ever interacted with a nipple.

How do you evaluate whether a $500 donation to a project that you know well is a good idea?

It just really depends on what the project is. If there were some generic way to evaluate all $500 donations, then some centralized organization would be doing that already. You have to use your own personal, human judgment.

Should we postpone AGI until we reach safety?

You can change the world, sure, but not by making a heartfelt appeal to the United Nations. You have to be thoughtful, which means you pick tactics with some chance of success. Appealing to stop AI work is out of the political realm of possibility right now.

What could one do with truly unlimited computational power?

The first algorithm I would use is this, to solve problems of mimicking a function with provided inputs and outputs:

For all possible programs of length less than X, run that program on the inputs for time Y. Then measure how close it comes to the outputs. The closest program is then your model.

This takes time O(Y*2^X) so it's impractical in the world we live in, but in this hypothetical world it would work pretty well. This only solves the "classification" or "modeling" type of machine learning problems, rather than reinforcement learning per se, but that seems pretty good to start.

For reinforcement learning, it just depends what you'd want to do in general. I would not just build a general AI and give it access to the internet, any more than I would bring an army of teenagers over to my house and give them access to my car and wallet. If you really had a super-powerful AI then I think the best way of increasing its practical capabilities over time while controlling it would be like any other technology - start a tech company, raise money, think of a business model, and just see what happens. That strategy seems way more likely that you could retain control over the technology and continue to express your own moral judgment over time. Compare to, for example, the scientists developing nuclear weapons, who quickly lost control to politicians. Maybe you could build a new search engine - that seems like it could be a lot better with real AI behind it.

Load More