Patrick Spencer — LessWrong

LESSWRONG
LW

Replying toLack of Social Grace is a Lack of Skill

Great points. I think a lot of people yearn for some ultimate form of communication that is unmoored to and unencumbered by social norms and Overton windows and emotions, and thus much more efficient at transmitting information.

But that form of communication is (a) simply inhuman, and (b) not that much more efficient.

Wrapping your ideas in a thin veneer of politeness is similar to encoding them in words. It's fuzzy and non-optimal, yes, but it's just how humans transmit information. It's not possible to fully mentally separate idea criticisms from personal attacks, even for those who try very hard to do that. So if you want to communicate more effectively, you put in... (read more)

Reason About Intelligence, Not AI

Patrick Spencer

3mo

I think that we reason too much about AI specifically, and not enough about intelligence more broadly.

This is a fairly obvious criticism. There are ways that biological and artificial intelligence are clearly different.

However, most AI discussion is largely upstream of biological-artificial differences, and so the discussion should be about intelligence, not AI.

Two areas where I believe this focus on AI is actively degrading reasoning are discussions of alignment and communication.

Alignment

Alignment is about agent-agent synchronized values, not human-AI synchronized values. The vast majority of agent-agent interactions that we have available to draw conclusions from are interactions where neither agent is AI. Instead they are between human-based entities, humans, other animals, or between two... (read 544 more words →)

Patrick Spencer3mo

I definitely think it's interesting that it's possible for N-D-substrate-computations to imagine / intuit N+1-D, but yeah, I feel like that's mostly a given because we have the concept of N+1-D in the first place.

There are different levels of "imagine / intuit" though. Some people have particularly good or bad intuition for the 3D space we live in. I took your claim to be something like "the average brain could intuit 4D just as well as 3D, maybe requiring slight modification". I think the modifications to reach true parity would be pretty extensive, because of how much 3D-specific architecture (as opposed to weights) human brains have. I do agree the modifications are theoretically possible, but the modifications to give a fruit fly human-level cognition are also theoretically possible with arbitrary modification.

Patrick Spencer3mo

This argument seems to work for N-D space for any N which doesn't seem right. I think we definitely do know less about 4D space than 3D, partly because we're much more interested in 3D, partly because there's just (a lot) more going on in 4D.

Intuitively it feels like current AI should be much better at learning navigation in 4D than human brains. Brains have real architecture-level, baked in task-specific circuits, which AI lack, and reconstructing a 3D world is arguably the most important of those. Sure, you could modify them with neurotech to change that, but you could do that for virtually any task so it doesn't seem very meaningful.

There's also... (read more)

Replying toLLM-generated text is not testimony

Patrick Spencer3mo

LLM-generated text is not testimony

It seems to me like both this post and discussion around Buck's post are less about LLM generated content and more about lying.

Opus giving a verifiable mathematical counterexample is clearly not lying. Saying "I think" is on somewhat shakier but mostly fine ground. LLMs saying things like "When I was a teenager" when not editing a human's account is clearly lying, and lying is bad no matter who does it, human or not. Extensively editing personal accounts indeed gets into very murky waters.

Patrick Spencer3mo

Skimming Starcloud's whitepaper: They say the radiators are "held at 20 C" without mentioning at all how they'd actually do that. For effective heat distribution you need fluid and small tubes, which are very prone to micrometeorite impacts, which means you need a lot more shielding mass. If you don't have effective heat distribution your chips get too hot.

Patrick Spencer3mo

This is what I was talking about, I should have been more clear. Year-average daily ground irradiance in the US seems to be about 4-5 kWh/m^2 (this includes night + weather + atmosphere effects, does not include panel efficiency) ~= 200 W/m^2. In space (assuming sun-synchronous or non-LEO, which are both more expensive than LEO) you get 1361 W/m^2. So about 7x. In the Sahara Desert it's ~300 W/m^2 so ground-solar worst case is 4-5x.

Patrick Spencer3mo

Yes, I've worked on spacecraft and you have to put about as much effort into dissipating your energy as collecting it. The benefits I've heard claimed for space based compute are minor at best, and the economic downsides of cooling are show-stopping before even considering the (many) other negatives. Electronics also have to run much hotter than otherwise if your only cooling method is radiation. I'm honestly struggling to understand why some people that I normally consider reasonable and realistic are giving it much consideration.

Patrick Spencer3mo*

I wish that article included any justification for that argument and I was unable to find any elsewhere by Chris Stott. Space energy today is certainly far more expensive than earth energy, probably 1000x.

Even as launch costs approach 0, things that can operate in space are more expensive than not. Vastly more difficult maintenance and cooling, radiation hardening, latency... I don't see how even 7x cheaper energy per area (current solar is ~200 W/m^2 avg, space maximum is 1360 W/m^2) can make up for those things. (edit: this accounts for weather + night + atmosphere effects, see below comment)

Patrick Spencer3moQuick Take

There is a recent intense interest in space-based datacenters.

I see almost no economic benefits to this in the next, say, 3 decades and see it as almost a recession indicator in itself.

However, it could allow the datacenter owners significantly less (software) scrutiny from regulators.

Are there any economic arguments I'm missing? Could the regulator angle be the real unstated benefit behind them?

Replying toOrigins and dangers of future AI capability denial

Patrick Spencer4mo

Origins and dangers of future AI capability denial

I'm not sure how to even engage with this category of claims because they totally ignore the obvious question that's automatically implied when you assert them, that being how exactly human intelligence differs from whatever term you use to say that LLMs don't have real intelligence.

Maybe belief in some of these terms is another good poll question. I suspect all of them are only going to get more common and dogmatic.

Origins and dangers of future AI capability denial

Patrick Spencer

4mo

In rationalist spheres, there's a fairly clear consensus that whatever AI's ultimate impact will be, it is at its core a capable technology that will have very large effects on the world.

In the "general public" sphere, things are very different. There's a less clear but real consensus that AI's ultimate impact will be negative, and not much agreement on the capability of AI as a technology.

I think it's very plausible that this could go the opposite way of rationalists, into a disbelief in the ability of AI as a technology, and eventually a more conspiratorial denial of even current AI capabilities. That is, the general public and policymakers could deny the existence... (read 2946 more words →)

•••

Patrick Spencer's Shortform

Patrick Spencer

4mo

This is a special post for quick takes (aka "shortform"). Only the owner can create top-level comments.

Dr Universe

Patrick Spencer

4mo

The Universe opened the door, walked into the small room, and closed the door. It adjusted its lab coat.

"Hello. I'm very sorry for the wait. We really had today overbooked, we're already an hour behind. Always more moral patients popping up, you know how it is these days." It walked over to the table and computer in one corner of the room and started typing something.

Humankind, sitting on the examination table: "Oh, that's alright, I was late anyways."

The Universe nodded but didn't respond, still typing. It finished up and wheeled its chair over to the exam table, clipboard in hand. "Okay Haman.. sorry, let me see here.. Humankind, what seems to be... (read 949 more words →)

I find that Claude is very bad at pushing back on the user's beliefs when there is any nuance involved. I just selected 3 random conversations that had this pattern:

Initial message from user asking neutral question about murky but factual topic - "would you say x thing is more like y or z?" not "how tall is the Eiffel tower?"
Model response giving evidence for a certain belief
Weak user pushback "but what about a and b?" where a and b are common, short talking points about this topic in opposition to the model's stated belief
Model immediately concedes and significantly retracts: "You're right to push back. I was conflating / being too hasty /

... (read more)