In lectures I thermostats as an example of an intelligent system that matches most attempts at defining intelligence. And I have no qualms saying they are intelligent. Intelligence is a very vague and very context dependent thing that can be at most used to compare some things in some scenarios.

A bimetallic strip is intelligent?

https://en.wikipedia.org/wiki/Bimetallic_strip

Thermal expansion is clearly a physical reaction; if arranged in an intelligent way, it's able to achieve some human-relevant objective... but if we're going to say that two pieces of bulk metal hooked together is intelligent, then the term "intelligence" is so broad as to be roughly meaningless.

A bimetallic strip is only intelligent when placed in the context of a thermostat system. Similarly, your brain isn't intelligent when removed from your body and placed on a table. Brains are only intelligent when they're in conjunction with living bodies.

A brain is intelligent regardless of where it is, so long as it's "powered" so to speak.

I used to think this but I don't know anymore. Without our bodies to provide stimuli what input is the brain going to use for thinking or reasoning? So much of what we do day to day is hard-linked to our body/sensorium. The brain without the spinal cord/nervous system is useless.

{{Citation needed}}

{{fart noises}}

Not without lungs, you don't...

But actually there's a real interesting thing here. People used to believe that the heart did the thinking. Phinneas Gage upended that theory, and now people think that the brain is the thinker. More recently, we're learning that the gut has a massive neural network, and the brain-gut connection is being investigated today. Likewise, "muscle memory" isn't in the muscles nor the brain, it's distributed among neural clusters all throughout your body. How much of "you" (your knowledge, your personality, your physical skills) is actually contained within the head is entirely non-obvious.

This is sophomoric. First of all, muscle memory is not contained in your muscles or "distributed" throughout your body but is known to reside in your cerebellum and basal ganglia. The nerves going down to your muscles are simply a conduit for actuation, a wire, nothing more.

As far as your gut, if you claim that a thermostat is intelligent then sure your gut is also intelligent (don't forget the microbiome in there too).

> This is sophomoric.

Says the user who just made fart noises. Your confidence here is dunning.

I literally just followed suit. Don't like the response, don't lead. You're being hypocritical.

Peripheral nervous system has significant plasticity and its reflex archs can produce complex and purposeful behavior without the brain. E.g. cats can walk with their brain removed and famously decapitated chicken do all sorts of stuff.

Octopi don't even have a centralized brain.

I'm not disputing that. What does that have to do with whether a human brain is inherently intelligent?

I was replying to this:

> First of all, muscle memory is not contained in your muscles or "distributed" throughout your body but is known to reside in your cerebellum and basal ganglia. The nerves going down to your muscles are simply a conduit for actuation, a wire, nothing more.

Sure, replace a living body with a machine, and replace the thermostat around the bimetallic strip with some other kind of mechanism. It makes no difference.

So a brain is only "intelligent" in certain (but infinite) contexts as opposed to intelligence as an inherent property of a brain?

If so, your definition of "intelligence" is incomplete. What are those contexts? Can you really generalize all of those contexts for all possible intelligence systems? That sounds like an epistemological nightmare.

Intelligence is not an inherent property of a brain, brains are only intelligent when placed in particular configurations with supporting hardware. Are you going to fart at me too?

Our brains are clearly physical reactions. Check and mate.

Perhaps check, but I don't think it's checkmate until you can explain the hard problem of consciousness via physical reactions.

What makes you assume thermostats aren't conscious?

Because they don't have any other higher level function that could reason about the state other than a single binary response. If intelligence is a continuum you might say it is at the lowest end (a bit state and actuation on that state). But consciousness seems more like a meta-intelligence that would require more than one bit and one actuation.

I don’t make any of those assumptions. My point was that I also don’t know that we can distill consciousness to strict materialism until we can find the mechanisms to explain the hard problem of consciousness.

this really is getting at something important. however the question I would pose is, if the thermostat is conscious, does it have a sense of self, i.e. an identifiable boundary between itself and the world around it? I would say not. My understanding is that humans have an identifiable region(s) of the brain that correspond to a sense of self: https://en.wikipedia.org/wiki/Neural_basis_of_self#:~:text=T....

I think in a lot of these conversations people say consciousness and assume that consciousness always mirrors their own, which just so happens to come along with a sense of self for 99.9% of sober individuals. So I think it's important to distinguish between the two things.

Would this imply that someone who claims to lose the sense of self (or seems to realize it as a hallucination) through “sober” techniques like meditation would no longer be conscious?

Quite the opposite. I'm saying that a sense of self is not necessary for consciousness. And that many people write off simple objects as having consciousness when really what they mean as that those objects don't have a sense of self.

Some definitions of intelligence (or even consciousness) are systems that simply process data. From that context, a thermostat is processing temperature data. It also underscores the GP's point that there isn't necessarily a consensus on the definition.

You can set a context as broad as you want, but it changes utility.

You can set a definition and context so broad that everything and anything is intelligent.

The possibility of an outlandish alternative, does not does inform on if a consensus exists or not

You can set the context as narrow as you want (e.g., abstract reasoning specifically defines intelligence). That might make your ontology feel crisp, but it also doesn't make it correct; it's still arbitrary and a matter of convention.

Indeed, the only value of categories in how useful they are. It's clear there's a continuous spectrum of intelligence/complexity. I believe the main reason we're still arguing in circles about what is or isn't intelligent, is that we also have a distinct, ill-defined category called "intelligence" that carries ethical / moral connotations. We don't know how to handle a continuum of moral weight / personhood.

It's useful in many ways to put a thermostat, an LLM, a monkey and a human on the same "intelligence"/"complexity" scale - there are many interesting things to discuss other than moral weight. And then, for discussing personhood, it makes no sense to include a thermostat.

sure. but there is value in defining a useful working convention so that people can discuss, learn, and study.

Interjecting the perspective that words are meaningless constructs is rarely helpful. If someone wants directions, telling them that there is no concrete metaphysical concepts of north, south, left, or right, isnt constructive.

I agree on the usefulness piece. But usefulness shouldn’t be conflated with the underlying truth. As the saying goes, all models are wrong, but some are useful. I can say that burying the dishes from a meal that made me sick may help avoid future food poisoning; while useful, it doesn’t get me closer to the underlying truth of germ theory.

I think the heart of these more philosophical debates is a desire for fundamental truth, not just a useful mental model.

I think intelligence is a continuum from not very intelligent to very intelligent. I guess a bimetallic strip would be at the bottom end of the scale above a lump of coal but below Einstein. At least in thermostat form it can keep something stable in contrast to say Turkey and it's currency as mentioned in another HN thread today. Maybe they could rig a bimetallic strip to raise or lower interest rates with a heat input driven somehow by consumer prices? They could then use that to replace Erdogan.

> thermostats as an example of an intelligent system

vs

> A bimetallic strip is intelligent

You're moving the goal posts

There are some who claim that a bimetallic strip shows evidence of intelligent design, but every fifth-grader knows that if a million volcanoes erupt over 5 billion years, there is a lot of weird stuff that can result from that random activity.

Language is quite flexible, but I don't think it is very helpful to call a thermostat intelligent. Doing so does not advance our understanding the sort of self-aware, theory-of-mind-holding intelligence that is generally meant by the term (putting aside its alternative meaning as a synonym for information), and adopting the broad definition just increases the burden of disambiguating what we mean when what we mean is the latter.

The broad definition is not justified by the fact that certain attempts to define the term would attribute intelligence to thermostats; a more useful conclusion to draw is that the definitions in question could be improved.

It is sometimes said that we won't be able to gain an understanding of intelligence until we have a precise definition, but it's actually the other way round: definitions get tighter as our understanding increases. We now understand many things that not only were lacking definitions in the past, but were not even imagined.

Thermostat with a servomechanism is self-aware of its state. A thermostat could be quite simply augmented to have a theory of another thermostat. Probably there are such coupled termostats.

I use intelligence mostly just as perspective of analysis to a system. E.g. Can it be seen to process information? How complicated is the transform from the information to some action? Does the processing lead to some good outcome, e.g. self-preservation of the system? Could the information processing be improved for this outcome?

Saying that a thermostat with a servomechanism is self-aware is just repeating the same broadening-beyond-usefulness with a different (though related) concept, in this case self-awareness. The difference between the self-awareness of humans and that which you see in a thermostat is vast, and understanding how a thermostat works gives no useful insight into how the human mind works.

Any system processes information, and in a thermostat it is not complicated at all. Even the most self-aware thermostats currently being manufactured are not going to improve themselves on their own volition. Applying the term 'intelligence' this broadly turns it into an empty tautology, or a little piece of uninformative noise on a communications channel, no more useful in understanding those systems than in understanding actual intelligence.

> I have no qualms saying [thermostats] are intelligent

> Intelligence is a very vague and very context dependent thing

That's either contradictory, to state that something has property X where X is vague to the point of undefined, or you are setting a very low bar for intelligence (I'm actually okay with the latter).

Just a question, if a bimetallic strip can be intelligent, is a marble intelligent for having the wit to find a low point on an uneven surface by rolling downwards?

The marble question is good. For most purposes I'd say the bimetallic strip (one connected to an actuator at least) is more intelligent than a marble. Studying dynamics of internal properties of the termostat is more fruitful than of the marble that is easy to describe as an unchanging object.

Taken to the logical conclusion, we are all just rolling downwards to a minimum energy state just like the marble.

I don't think intelligence is an intrinsic, let alone binary, property of an object or a system.

Didn't Nest have a >$1B valuation at some point? /s

Juicero was $120M /not s

> I'm not sure I can subscribe to this definition...

Is this because it is a completely man-made system and not one that evolved slowly over time through natural processes?

My thermostat may not be intelligent but it is certainly smart... At least it says so on the box.

Anyway, the strip does contain the state of the world around it: the temperature is modeled by how much the bimetal is bent. I think indeed it is a minimal example of a complex system, one that at first glance defies explanation, it seems to have purpose (keep temp stable), until you understand the inside.

Anyway, "Is a virus alive?", "Are these specimens the same species?", ... Us humans like our boxes, but at the edges they almost always go wrong.

A thermostat is certainly intelligent.

The definition of intelligent I give is “to mitigate uncertainty.” If it does not mitigate uncertainty, it is not intelligent.

It is merely of constrained intelligence. Perhaps your expectations are too broad.

If the thermostat reacts appropriately to environmental changes then it is performing its role intelligently.

In the context of a person being trapped in a cave, is some aspect of light shining distantly intelligent? It's reducing uncertainty about where a viable exit is, after all.

Information is the reduction of uncertainty. If it does not reduce uncertainty it is not information.

Intelligence is the mitigation of uncertainty.

The revelation of the light is information. Navigating the contours of darkness is intelligence.

If the cave walls glowed in response to proximity (of the observer) then the light source would have a modicum of intelligence.

But what if the cave walls glowing misleads the observer and they get lost further? Then it wouldn't seem to be intelligent. In any case, the light source being intelligent occurs in the context that there is an intelligent observer who may be able to escape easier given a light source. The observer sees the light, evaluates it, and makes a decision. Even if it's just instincts, there is a decision made. What might be intelligence is if the cave (or elements of it) acts according to "this person is lost at this location, so light up this specific pathway". If the cave somehow facilitated more functionality, such as leading a person to clean groundwater, it would be more intelligent. In the context of helping the person escape or aiding them in the meantime, the cave tends towards accidental sophisticated technology ("...[which] is indistinguishable from magic"). Functional intelligence, I'll call it. Then perhaps there could be a theory of intelligent agents, who have goals and plans.

It's not just the internal state but the prediction that makes it intelligent.

Your brain is taking in a lot of information at the edges of your awareness, light, sounds, touch, etc. are all getting absorbed and transmitted to your brain. As that information is transmitted along your neurons it's getting summarized, then merged with other summarized information and summarized again. The brain is getting summaries of summaries, and developing a unified categorizing of the global state across all it's inputs.

Then the brain takes that summary and makes a prediction about the future state. The summarization is energy-efficient. By categorizing all that data into a global state you make decision making possible. "When my boss seems stressed all week, then calls a bunch of people one-by-one into his office on Friday afternoon, I know lay-offs are coming. I better polish up my resume." From "stress/anxiety/unease" in the environment to "danger is coming I need to fight/flight".

Your brain is taking that summary/categorization and figuring out what it needs to do next. If "X" happens then I should do "Y" to "stay-safe/get-food/find-a-mate". The brain is very good at capturing and summarizing data, and making a prediction because that process is much more efficient than doing otherwise. Instead of foraging everywhere for food and hoping I just bump into something that will provide sustenance, I know if X, Y, and Z happen then food will be "here", and I can get lots of it.

You can apply this same model to all actions the brain directs. It also helps make sense of why maladaptive behaviors develop. Sometimes the summary is incorrect, or was formed based on past information that no longer applies, and it may need to be unlearned.