What is it like to be a thermostat? (1996)
annakaharris.com
annakaharris.com
This ruminating has led me to believe that time and logic are human concepts, and are not as universal as is commonly believed. With recent insights into neural networks, I wouldn't even be surprised if the laws of physics follow from the way our brains are wired, instead of the other way around. Perhaps this is simply a modern take on idealism or realism, but I can't find a strand of philosophy with which I feel at home.
Obviously, there is a bootstrapping problem with trying to reason from something that cannot reason. And I am well aware that my brain must exist in some form of reality. To conclusively prove some apparatus for that is way out of the scope of science. Scientifically there is probably very little to learn from this anyway, apart from opening one's mind to some alternative possibilities. It's a fun exercise, though.
However, the entire discussion about what consciousness is, strikes me as less interesting. Is this really more than being able to conjure up memories of past experiences?
[1] https://en.wikipedia.org/wiki/What_Is_It_Like_to_Be_a_Bat%3F
I don't think memory and consciousness are intrinsically linked. Memory is something consciousness can be aware of, but it's not consciousness itself. Someone can have their ability to process and remember memories permanently or temporarily damaged, and yet still have a conscious experience. An AI can have memory, but not have a conscious experience. (Or at least it seems that way - if something like Integrated Information Theory is true, then maybe AI does have some sort of first person conscious experience. I tend to doubt that, but I could be wrong.)
EDIT: although I might be conflating short term and long term memory. I wonder if consciousness requires at LEAST some form of memory processing, even if it's just the past couple of seconds. Perhaps the "Strange Loop" needs at least that to arise. I'm not sure.
There is no atemporal “frozen” state we can call consciousness. It is dynamic.
That is what bothers me about the Chalmer’s piece. There are not three states of a thermostat’s “What’s it like”. He is showboating his writting chops and parroting Nagel’s dualism.
That is a huge assumption we just have no way of knowing. For starters, we don't know whether we are in an atemporal "frozen" state or not.
My take: Epistemology is metaphysical bs, and “truth” is a convention we agree to within communities of speakers.
It's reasonable to assume, and so allow us to talk about the passage of time even though we have no conclusive evidence for it, but that does not make it reasonable to claim conscioussness requires a dynamic system, because we don't know and can't know, only act as if it does.
Panpsychic "everything is alive" is better than "nothing else is conscious because I'm not them", but only by one degree.
> Where does it hurt?
Pokes leg, ow
Pokes arm, ow
Pokes stomach, ow
> Everywhere!
> I think you sprained your finger
From the article, emphasis mine: "But we should not be looking for a homunculus in physical systems to serve as a subject. *The subject is the whole system*, or better, is associated with the system in the way that a subject is associated with a brain."
Is this dualism because it retains the idea of the subject at all? But Chalmers states "the subject is the system", and so seems to reject the notion of mind/body duality. I haven't read about this in much depth, so I don't have a very sophisticated understanding here.
For sci-fi, have a look at Blindsight, by Peter Watts (for free on his website: Https://www.rifters.com/real/Blindsight.html)
I have read Thomas Metzinger's Ego Tunnel (2009), but as far as I understand it, he takes on a naturalist standpoint, and assumes that consciousness arises from that.
I prefer to take a radical agnostic point of view, where consciousness does not even have meaning outside of those who experience it. Implying that "meaning" or "reasoning" make no sense universally.
Unless I'm missing something, perhaps you care to elaborate?
Mindware by Andy Clark is also a great book on these topics.
https://global.oup.com/academic/product/mindware-97801998281...
That is how I perceive meditation to be. At least, the end goal that I have yet to achieve, anyway.
Mindfulness practice is toward the other side of the spectrum: The goal is not to suppress your own sensations or thoughts, but to be mindful of them and observe them in a detached manner. The best analogy I've come across is to sit at the side of the river and watch the boats go past, instead of jumping on them and racing down. But you're not trying to clear the river of boats.
There's overlap, in that the finer control you want to have over your ability to be mindful of specific aspects of your thoughts, emotions, body etc., the more concentration you need to be able to muster to calm yourself enough.
Your last comment strikes me as strange for someone who seems to be well read on the topic. Saying that consciousness is an ability to recall memories doesn’t really describe what it is in the natural sense. The memories themselves are composed of conscious experiences, so that definition is circular. An explanation of what consciousness is would include an explanation about why, for example, chocolate tastes the way it does, rather than like vanilla, or some other completely unknown taste. Until we can explain its character (rather than just describe it), we can’t explain what it is. It’s sort of like dark energy: we can describe the phenomenon but we haven’t fully explained what it is.
A sister comment suggests that memory and consciousness are less intrinsically linked that I tend to believe. It might be fruitful to come up with a decision tree on what people believe consciousness to be :)
Personally, apart from the rock meditation, I am not as much interested in the definition of consciousness, because I think it is a trap, based on a categorical mistake. I'd rather get away from the anthropocentric viewpoint. Then again, I sometimes doubt that a scientific method will get us there.
To mess with your head a bit more:
We know of no other way that we know the flow of time than indirectly through memory of the flow of events and sensory inputs.
And so while it seems probable that our brains must exist, consider that e.g. a simulated mind that is paused, and where the same step is executed over and over, with no ability to change state, would have no way of telling that apart from the same step being executed only once, to move on to the next.
In that case it's not clear that there'd need to be any full brain, just whatever consciousness is, in a state that has a notion that it has a memory of a past instant.
Put another way: Your consciousness could be just a single frame of a movie, hinting at a past and future that might not exist.
Forever repeating the same infinitely short instant, and nothing else. Maybe the universe is just a large tableau of conscious instants statically laid out and never changing or interacting with each other. We wouldn't know any different.
Of course that is entirely untestable, and so just a fun philosophical question to ponder, mostly as a means to point out just how little we can know, and so how willing we need to be to accept that what matters isn't what we can know, but what is likely to be able to affect our observed world.
E.g. I see myself as a materialist (philosophically speaking) not because I believe it is or can be proven, but because it is our observable reality. If that materialist reality is shaped by our minds and only exists in some abstract form, or we're all in a simulation etc., then that is irrelevant unless/until we find a way to pierce the veil and/or those aspects "leaks" into our observable reality somehow.
You have to give it a chance. He is first building up an argument about why consciousness cannot depend just on the physical substrate itself but rather on the "interpretation" of this. It is very important to understand this part/argument. What follows is something that resonates with you namely how our consciousness is now 'tuned' to the current physical laws.
in a very real sense, we share this
"Moving down the scale through lizards and fish to slugs, similar considerations apply. There does not seem to be much reason to suppose that phenomenology should wink out while a reasonably complex perceptual psychology persists… As we move along the scale from fish and slugs through simple neural networks all the way to thermostats, where should consciousness wink out?"
The author seems to have succeeded in answering her own question (at least in hand-wavy fashion) at the same time as posing it, as well as implicitly defining consciousness. So, yeah, it's not like anything to be a thermostat.
By the way: The headline is, I assume, by Annaka Harris, while the essay is by David Chalmers.
Sure - there is a difference between merely being cold and also being aware of being cold. A piece of ice (or a cold thermometer for that matter) can be cold, but to experience being cold - to be aware/conscious of it - requires some minimal level of cognitive apparatus ("a reasonably complex perceptual psychology") to process those sensory inputs and contrast them to the differing sensation of not being cold and be able to think about it!
There is some evidence, such as "blindsight" (ability to see without being aware of it - a loss of visual consciousness) that consciousness may not only require the mental apparatus to process a sensory input, but may also require specific neural pathways (which may be missing, or damaged) to gain access to specific internal neural state in the first place.
It's difficult to know exactly where the line is - which animals do have a sufficiently complex brain to able to introspect on and experience their own state, but clearly simple neural nets (e.g. anything without feedback paths) don't, and simple animals like insects don't either.
https://www.youtube.com/watch?v=PnQ9zkBzbYc
EDIT: there is of course a bit about thermostats: https://www.youtube.com/watch?v=PnQ9zkBzbYc&t=1137
Then, I know people with autistic kids. Wow.
This applies to so many things, not just health. You only appreciate it when it happens to you, or you get a snapshot of someone else’s life and it humbles you.
A CGM and insulin pump have made my life easier.
How old is your son?
The mental burden is SO HIGH for both the person with the condition and their caregivers. At some point if it feels right, it might be worth having him see a therapist or a counselor who specializes in chronic diseases. There's a very high correlation between depression and type 1 diabetes (last I read, you're like 300% more likely to experience severe depression as a T1D.) Especially once he goes into his teenage years, that's a really hard time for a lot of us diabetics (and sometimes during the "rebellious" phase, a kid will intentionally stop taking care of their diabetes).
> His pump beeps at him so often he's just become accustomed to it and ignores it
This is a real problem with insulin pumps and CGMs, because alarm fatigue is a real thing. It makes me mad that they insist on putting in all these alarms you can't configure, because eventually you just start ignoring it. For my CGM, I use xDrip which gives me much more customization around my alarms: I just leave on the ones that are important to me.
In a way it's eliminating too much mental effort; while it's useful as a backup, the fact I sometimes completely forget to take insulin with meals is not ideal, even if the closed loop notices and takes care of it for me (since there's inherently more lag when relying on the loop than if I dialed in the insulin at meal time).
I have the Omnipod 5 which is decent, but it's "closed loop" abilities are extremely conservative and a bit disappointing. I've considered going the full way with androidaps, but haven't taken the leap yet.
Are you using a Dexcom with your omnipod? My impression is that the quality of sensor is the most important factor; with a flaky sensor looping algorithms will be more conservative since they (reasonably) prioritize avoiding lows.
Nice!
> I enter carbs but only to a resolution of 10g; and as I mentioned earlier I sometimes forget to enter them entirely.
I'm pretty bad at accurately entering carbs. I eat out enough that I've just gotten to the point where I just come up with a rough estimate (which is often wrong). Which means I have to adjust later pretty frequently with snacks or correction doses. Not ideal, but I got so much "carb counting fatigue" over the years.
> Are you using a Dexcom with your omnipod? My impression is that the quality of sensor is the most important factor; with a flaky sensor looping algorithms will be more conservative since they (reasonably) prioritize avoiding lows.
Yeah, I'm on the G6 still. I think the quality of my readings is pretty good, but the Omnipod 5 algorithm is pretty conservative anyway: rather than naturally bringing you down from a high, you usually have to explicitly give yourself a correction dose. Which I suppose makes sense, but I have read that some people have had a lot more success with looping in Androidaps vs the Omnipod 5's algo.
I also wish you could set the "target" a bit lower in the 5. You can't tell it to target below 120: I feel like 100 or 110 would be more reasonable, but again, it makes sense that they prefer being conservative.
I live in an area where its cold at night and hot during the day and I am bad at remembering to change the thermostat from mode to mode and haven't found a programmable IOT thermostat I can write a script for, recommendations welcome!
A bacterium finds food with a simple set of states. At it's most basic:
- you are experiencing an increasing concentration of food. you keep swimming straight ahead.
- you are experiencing a decreasing concentration of food. you move in a continuously randomized direction.
this eventually gets them onto a track where they are moving towards food.
Extremely simple like a thermostat, yet effective.
I definitely take issue with Chalmers' opening sentence of his final paragraph: "… A final consideration in favor of simple systems having experience: if experience is truly a fundamental property, it seems natural for it to be widespread." I feel he is putting the cart before the horse here - something that seems quite common in the philosophy of mind - by first deciding that experience is a fundamental property, and then using it to justify the assumption that it is widespread. This strikes me as almost circular, as it seems one could at least as reasonably justify it being fundamental on account of the arguments for it being ubiquitous.
This is one reason the topic is so slippery.
I.e. I don't think thermostats want anything. They don't have a capacity to care whether they fall apart or not, whether anyone is satisfied with their function or not. But, life, even in the very simplest form wants something. Experiences to living organisms is what makes them more effective at doing what they want.
What makes living creatures want something: I have no idea. I remember hearing a hypothesis tying this to better preservation of energy... but I don't know about that. But, if I had to guess, it must be something that's available to micro-organisms, so, it has to have some physical / chemical explanation...
Still that boils down to: survival has to be a property (or even the definition?) of life.
So maybe the question of life introduction is, when did survivorship emerge?
Humbert Maturana and Francesco Valera(1979) Autopoiesis and Cognition: The Realization of the Living. ISBN 90-277-1015-5.
Not an easy book, but one that answers your question.
Terry Winograd loves this book. I am a neuroscientist and only found this gem late in my career. Damn!
> phenomenal: Known or derived through the senses rather than through the mind.
> phenomenology: A philosophy or method of inquiry based on the premise that reality consists of objects and events as they are perceived or understood in human consciousness and not of anything independent of human consciousness.
So the claim (highlighted especially in paragraph 3) is that, outside of humans, things that are perceived may also exist in conscious thought (of non-humans).
In contrast if you want modern post-cartesian scientific thought on consciousness then hit Dennett hard for philosophy or Ray W. Guillery if you want hard neuroscience (The Brain as a Tool).
Instead try asking yourself "what is it like to be asleep?" or "what is it like to be waking up?" or "what is it like to be heavily sedated?"
We all experience various gradients of consciousness every day as we do things like drift off to sleep or slowly gain consciousness in the morning. You don't have to try to imagine the experience of another primitive life form when you can just recall what there is or isn't to your own conscious experience as you drift between states.
Because yes, you acknowledge "experience", but you make it a function of a physical state described in such and such a way. In the same way that a set of particles at points A, B, C, .. correspond to such and such a (e.g. electric) field strength at point Z, we now imagine it could correspond also to such and such an experience.
It's just barely "experience" on its own terms. and elicits a kind of epiphenomenalism and powerlesness. The thermostat*, after all, doesn't choose anything nor does it profess to have any agency. So agency ought to end up some kind of ephiphenomenal "observable" of a system.
But besides being deflationary in this distasteful way, what bothers me with pictures like this is that they make use of entirely subject-made divisions between objects and their environments, and presume that they might correspond to experiences because - why not? Why not thing of the bottom and upper half of thermostat as corresponding to two fields of experience? Or the quarters, sixtheents, and so on until we get to individual atoms.
The thermostat doesn't "care" if I think of it as the wax and glass separately, or as a single object containing both. But we do have a unified field of experience, and it doesn't matter how another person "cuts us up" in their mind, whether it is as atoms interacting, organs behaving in unison, or just as a "body".
It seems silly to say that between me and Bob having our separate experiences, there is an experience corresponding to "me and Bob", supposedly free-floating somewhere just by virtue of the two of us being cognizable as a physical system.
It turns "experience" and that infamous "qualia" from something that's the most direct and obvious to a weird phantom as the output of a presumed equation which maps some description of a physical state to an "experience".
No wonder you'll find people who'll retort that they don't experience things or that their consciousness is illusory - they have these weird detached notions of experiences to fight against.
* I imagined a thermometer throughout reading this piece, hence the mention of wax and such. It doesn't really change the point so I'm leaving it.
This leads to two pounts of view - scientific, leading to reduction and more philosophic - there's no way to describe it since it is _super-natural_.
I lean to the more scientific approach, we are not more and not less than the sum of our parts and, each of our parts, at any sub-scale, has some resemblance to a thermostat: some object that reacts on its environment.
Nonetheless, it is far from compelling even as a "weak-problem" hypothesis and is an abstract angels-on-hairpin musing that truly puts experience outside of the bounds of investigation. Because, after all, if experience is an empty epiphenomenon which exists for any, anyhow-delineated physical system out there, where does that get us? We've made an assumption we cannot prod scientifically, yet it hides behind the scientific veneer of reductivism.
https://en.m.wikipedia.org/wiki/Humberto_Maturana
For Hofstadter consciousness requires some form of recursion—what he calls strange loops. Our brains are recursion machines “we” can partly control “ourselves”.
For that reason, it isn't 'like' anything to be, say, most insects, let alone a thermostat.
But I have this question- why am I interested in the date? What changes in this message when I know the date?
I feel like knowing the date sets up a conflict between "this is out of date" vs. "this is before the era of mass garbage generation" (more generously described as "with time only the classics survive)
This kind of showboating in 1996 is what allowed him to achieve his professional goal—-to be the modern voice of archaic Cartesian dualism.
There actual is no hard problem. Ask neuroscientist “What is the hard problem” and you will get blank stares. They just do not know, or like me, do not care. It is a muddy residue of Cartesian dualism. Boring.
https://aeon.co/essays/as-real-as-it-ever-gets-dennetts-conc...