Book Review: We Are Electric by Sally Adee, Allen&Unwin 2023

[Jonathan Edwards]

Humans perceive it as taste, love, pain or happiness when all that's happening is certain portion of the brain light up.

Which of course is in physics terms a completely meaningless or unexplained account of things. We need physics to account for tastes and colours. We need that because, as Eddington (the guy who first confirmed Einstein's general relativity) pointed out, we actually know nothing about the nature the physical other than its dispositions to trigger colours, tastes, sounds etc. Physics is defined in terms of qualia. Something that has been understood by the people who invented physics (Galileo, Descartes, Newton, Leibniz, Maxwell, Mach...) for centuries but got lost in dumbed down schoolroom teaching in the twentieth century.

 

[Eddie]

Which of course is in physics terms a completely meaningless or unexplained account of things. We need physics to account for tastes and colours. We need that because, as Eddington (the guy who first confirmed Einstein's general relativity) pointed out, we actually know nothing about the nature the physical other than its dispositions to trigger colours, tastes, sounds etc. Physics is defined in terms of qualia. Something that has been understood by the people who invented physics (Galileo, Descartes, Newton, Leibniz, Maxwell, Mach...) for centuries but got lost in dumbed down schoolroom teaching in the twentieth century.

If you showed someone from a thousand years ago that bits of rock can be organized in a way to carry out complex programs and answer all sorts of problems, they would probably think you had magic powers. Even now, we don't know how a computer runs complex programs, how all the individual transistors were switched on or off to achieve the outcome. But it works, even if we don't know exactly what every transistor is doing at all times. And while human experience is clearly much more complex, I don't see how qualia helps explain what is going on. Do you think there is something outside of matter and energy interacting that produces tastes and feelings? If not, then mustn't qualia map onto physical processes in the brain?

 

[Utsikt]

Even now, we don't know how a computer runs complex programs, how all the individual transistors were switched on or off to achieve the outcome.

I don’t understand this claim. We know how computers work at a fundamental level, and programs can be optimized for this if we want to. It just really isn’t worth the effort unless you’re making a supercomputer for a very specific purpose.

What we usually do is to optimize the hardware, like how modern chips are much better at running AI than previous generations.

 

[Jonathan Edwards]

I don't see how qualia helps explain what is going on. Do you think there is something outside of matter and energy interacting that produces tastes and feelings? If not, then mustn't qualia map onto physical processes in the brain?

Right. Qualia do not help explain what is going on by adding something extra to the physics account. 'Matter' is probably not a very good term for what the world is made of these days. Billiard balls have been replaced by fields that evolve through spacetime by events of excitation. All that we have are patterns of events. But let's call it matter and no we don't want anything more than that, with some other rules, apart from anything because when we find those rules we will have to call it matter as well.

So yes, exactly, qualia should map onto physical processes or events.

A popular way of looking at physics now is to say that it all consists of events of informing - that 'matter' is really the passing round of information. There are complexities to that but I think it stands up fairly well. When we experience a sunset something must be informing something else. So the obvious conclusion (which was indeed obvious 300 years ago) is that qualia are just what it is like for the second something to be informed by the first (maybe one is a field and the other an excitation of a coupled field). We could say that in rocks and computers informing goes on but it isn't like anything to anything to be informed but why not? The only informing we have direct knowledge of is the informing that is like something - a bunch of qualia.

As science is at present qualia aren't much use to us for predicting outputs because they are only half of what goes in to an event of being informed. The other half is the nature of the thing being informed. To predict the result of a negative electrical potential informing something we need to know if that something is negatively or positively charge. Since we know absolutely nothing about what the 'me' is that is informed by signals looking like a sunset we cannot predict just on the basis of the qualia involved.

The reason why the qualia of the sunset are relevant though is that they form a rich pattern that represents an outside scene and so it seems that something is being informed richly an d in a way in which the richness provides detailed representational data. In a computer nothing like that happens. We know that because we have designed them to work with computational events in which two inputs just represent 0 or 1 (if they even represent that). The rich representational patterns ofqualia we have access to mean that we have to look for events in brains where rich patterns of signals that represent outside events contribute to the same event of informing.

That turns out to be easy enough because each individual nerve cell gets rich patterns of signals that represent the outside world all the time. The difficult questions are what part of the nerve cell (what discrete dynamic unit carrying a mode of excitation) is getting the informing directly and what is the code for sunsets and tomatoes and trumpet sounds laid out in the synapses.

 

[Eddie]

I don’t understand this claim. We know how computers work at a fundamental level, and programs can be optimized for this if we want to. It just really isn’t worth the effort unless you’re making a supercomputer for a very specific purpose.

What we usually do is to optimize the hardware, like how modern chips are much better at running AI than previous generations.

We know how computers work at the fundamental level. But as programs become more complex, it becomes more and more difficult to know how the 1 and 0s comes together to produce the desired result. In programs that use large data sets, I think it would be almost impossible to go back and figure out exactly how binary is used to get to the desired result. My point was that just because we don't know all of the steps to achieve the desired outcome in computing, that doesn't mean there is something other than "matter" going.

This doesn't directly translate to biology, and even less so to qualia. However, in a very basic organism we can have a general sense of how the the various inputs produce the intended result. As we scale up in complexity, it becomes more difficult to connect inputs and outcomes. It also is the case that this complexity seems to correlate with how conscious an organism is. So I think that qualia must be a result of complex inputs even if it would be incredibly difficult to know all these inputs.

 

[Jonathan Edwards]

It also is the case that this complexity seems to correlate with how conscious an organism is.

Do we know anything about that? More complex organisms appear to be able to form representations based on more complex inferences - such as classes possible future events - but is that consciousness rather than intelligence? I don't think we know anything about the richness of conscious events in any organism other than ourselves - with a reasonable extrapolation to other humans. And for ourselves we only talk about events connected up to reporting circuits that monitor content. Our cerebellums may support much richer consciousness in the Purkinje cells as far as we know.

So I think that qualia must be a result of complex inputs even if it would be incredibly difficult to know all these inputs.

Right, but inputs at the point of the event of experience rather than sense organs. We do not know a lot of detail but often in biology we manage to generalise. We have now got to the point where there is a strong consensus that dendritic trees in at least some pyramidal cells get inputs of 'expected' scenarios at one end and 'actual scenarios' at the other. We can know where the inputs come from for at least some synapses on sampled cells.

 

[Eddie]

Thanks for the previous post, I thought it was very interesting. I just had a few questions; why do we think everything consists of the transfer of information and not say the transfer of energy?

If everything consists of informing, I'm not sure why the concept of qualia seems to be unique to brains. In the case of a rock, there is still the matter to be informed. And if the rock "sees" a sunset, then the atoms of that rock will have a marginally different energy as a result of experiencing that sunset. So the information in some sense is still transferred, in humans it is just transferred in other ways too (just as using chemical energy from food to causes events in the brain to store that event) because storing such events is useful to our survival.

If computers had access to more inputs could they potentially experience qualia in the same way a person does?

 

[poetinsf]

A computer can simulate pattern recognition in that it can produce appropriate outputs that are informed by whole patterns but it does that by interrogating individual bits of the pattern and using computational manoeuvres that can derive a mathematically correct response from a series of such one bit interrogations.

How do you think the brain does it differently? It interrogates sensory input signals from, say photoreceptors, outputs recognition signals, say "red". The output signal is then fed into various other circuitries that invokes various meanings attached to "red", therefore creating the illusion called "qualia". The world is a simulation in the brain, of the programmer's or yours. There is no such thing as "pattern" that the signals stand for. The signals *is* the pattern.

So the brain's 'circuitry' is not the same in principle at all.

Mathematics says otherwise. Synaptic connections strengthens/weakens according to the training and that's what learning is. And that is why the modern AI is called neural, after all, because it strengthens/weakens the connection by adjusting the dial after each training.

The other is the quale red, which is a manifest property of an event deep in the brain that has nothing whatever to do with the previous dispositional properties as far as we know.

Yes, they are various meanings attached to "red" pattern. And they are also learned through previous training and then attached to the "red" pattern, to be invoked when "red" is recognized. Some people mistake the linkage as some sort of independent entity and call it qualia. It's not. It's only a learning.

Not scientific? Well, it was hard science to Sir Colin Blakemore, one time head of the UK Medical Research Council and someone I discussed perception with often in our shared neuroscience and philosophy seminars at University of London Senate House.

I'm not familiar with him, but I'd be happy to argue with him too on the topic of qualia vs information only universe!

 

[Jonathan Edwards]

why do we think everything consists of the transfer of information and not say the transfer of energy?

Energy is just one of several aspects that get traded in field interactions. Energy is a scalar and the information being passed can be complex domains of vast numbers of vector values (notionally inifinite numbers but that is probably a red herring). There is a huge amount to say behind all this. The whole idea of 'particles' interacting really has to go. Interactions involve new excitations arising to replace old ones in the context of coupled field patterns.

This is a quantum level view but I think aquaria have to be at a quantum level simply because the quantum level is now what we use for direct interactions between indivisible dynamic units and anything other than that raises problems of causal regress, baulking and combination (sorry there is a huge amount of background to all this). It is likely to involve higher order quantum modes in condensed matter but that is easy enough because sound waves qualify for that.

I'm not sure why the concept of qualia seems to be unique to brains

It isn't. A lot of major natural scientists over the years have been panpsychist. It seemed to me to be obvious position to take when I first got interest in this 60 years ago. There are a lot of philosophers who think that experience and qualia must be limited to higher organisms but I think that is just a hangover from Abrahamic religious teaching. And of course it would not be the brain that 'has qualia'. It would be individual events somewhere in the brain. Similarly it will not be the rock that sees but modes of excitation that inhabit the rock that undergo fundamental direct interactions. And the same in computer gates. An electron field mode such as a semi-free valency mode in a semiconductor shifting to a new type of mode under the influence of a voltage might be a candidate.

If computers had access to more inputs could they potentially experience qualia in the same way a person does?

Again it would be events in computers have access to more inputs and specifically inputs that usefully encoded representations.

Things do get very complicated if one goes further (I have published on all this) but I might just add that there are interesting reasons for thinking that neurons are the shape they are and have the skeleton they do precisely because it allows them to support modes of field excitation that will be pretty unusual elsewhere. Roughly similar modes may occur in a lot of ordered condensed matter structures but they will probably be useless for supporting anything we would conducer meaningful because you need all sorts of parameters to be matched for that to be feasible.

 

[Eddie]

Do we know anything about that? More complex organisms appear to be able to form representations based on more complex inferences - such as classes possible future events - but is that consciousness rather than intelligence? I don't think we know anything about the richness of conscious events in any organism other than ourselves - with a reasonable extrapolation to other humans. And for ourselves we only talk about events connected up to reporting circuits that monitor content. Our cerebellums may support much richer consciousness in the Purkinje cells as far as we know.

I certainly don't, as I am not a biologist. I tend to assume that many other animals have some level of conscious experience though. I would argue that if an organism seems to react in a similar way to pain as a human does, then we should assume they are in some way experiencing that. It does seem that the more complex the organism, the more it has human like responses to stimuli and thus probably more conscious. Especially because I believe consciousness is an evolutionary adaptation, I think that animals more closely related to us are probably more conscious.

Also it is a bit of a slippery slope as to where you draw the line as you can always say we shouldn't extend that assumption to other apes, or even other humans. And even if we knew exactly how the conscious experience came about, that doesn't necessary mean that others are actually experiencing it. Everyone else could be a simulation and your own experience is the only "real" one.

 

[Utsikt]

But as programs become more complex, it becomes more and more difficult to know how the 1 and 0s comes together to produce the desired result.

It becomes more difficult because it would take an unreasonable amount of resources to look at it. But a computer is deterministic and based on the laws of physics, so it works exactly the way it’s instructed to work.

 

[poetinsf]

Which of course is in physics terms a completely meaningless or unexplained account of things. We need physics to account for tastes and colours. We need that because, as Eddington (the guy who first confirmed Einstein's general relativity) pointed out, we actually know nothing about the nature the physical other than its dispositions to trigger colours, tastes, sounds etc.

I think this is where you and I disagree. Science is not there to explain or account for things, as if there is such a thing as things. It is there to predict. It only needs to predict red, as defined with its wavelength, means certain temperature, for example. We can leave the explaining or qualitative meaning to philosophy and religion.

Physics is defined in terms of qualia.

You surely mean observation and measurement, not qualia. Physics deals with the wavelength, not qualitative meaning of red, or quale red as you put it.

 

[Jonathan Edwards]

How do you think the brain does it differently?

The difference is that the integrator units respond to patterns of up to 100 maybe even a thousand inputs arriving in up to 50,000 input channels. Those patterns represent dynamic patterns in an outside world that we only know through the inner patterns. I am not sure in what sense this would be an illusion. It is knowing through the only route available. The world is not a simulation in the brain. The world is a network of events that one part of the brain represents for the benefit of another which is primed to respond in useful ways.

Synaptic connections strengthens/weakens according to the training and that's what learning is. And that is why the modern AI is called neural, after all, because it strengthens/weakens the connection by adjusting the dial after each training.

But that is just metaphor. There are no 'weightings of synapses' in these machines, just routines that generate outputs as if there were, through vastly long winded Turing machine sidestep manoeuvres. If you look at the circuitry of one of these AI chatbots you will find not a hint of an integrator with 50,000 input channels.

Some people mistake the linkage as some sort of independent entity and call it qualia. It's not. It's only a learning.

This has been a very popular view espoused by people like Daniel Dennett who claimed to be a defender of scientific materialism. Dan got it from Gilbert Ryle in the behaviourist days. Trouble is Dan didn't actually know anything about perception physiology and nor do all the AI people. Red is not just learning. It is an experience. If there are no relevant events of experience in your head and you are a 'zombie' then I cannot expect to persuade you but it would be pretty unlikely.

Is pain just learning? Or is it painful as well? Mine is painful? Why are people with ME/CFS grumbling if pain is just learning. Sounds seriously BPS to me, seriously behaviourist. Maybe I am wasting my time worrying about what PEM is.

I'm not familiar with him, but I'd be happy to argue with him too on the topic of qualia vs information only universe!

Sadly both Colin and Horace are now gone.

 

[Eddie]

It becomes more difficult because it would take an unreasonable amount of resources to look at it. But a computer is deterministic and based on the laws of physics, so it works exactly the way it’s instructed to work.

Exactly, and I think consciousness is deterministic and based of the laws of physics. I think the main reason why many people think differently now is because the amount of resources we would need to know all the inputs is so many times greater than with a computer.

 

[EndME]

Exactly, and I think consciousness is deterministic and based of the laws of physics.

If it's based on the laws of physics then I suppose it can almost definitely not be deterministic.

 

[Utsikt]

I think the main reason why many people think differently now is because the amount of resources we would need to know all the inputs is so many times greater than with a computer.

The largest computer chips have trillions of transistors, and the largest AI models have trillions of nodes. I believe the brain has ~100 billion neurons and an estimated ~100 trillion connections.

It does seem like scale or complexity creates thoughts. I don’t even know if it’s required.

 

[EndME]

If it's based on the laws of physics then I suppose it can almost definitely not be deterministic.

Computers already have absurd amount of computational powers. There seem to be other more intractable details they miss, call it dynamic topological network effects for the lack of a better word.

 

[Utsikt]

If it's based on the laws of physics then I suppose it can almost definitely not be deterministic.

Depends on the scale.

 

[EndME]

It does seem like scale or complexity creates thoughts. I don’t even know if it’s required.

Who says multiplying matrices sufficiently often (and applying one non-linear operation after each step) is enough to be complex? My understanding is that all "AI models" currently lack complexity and largely excel on the computational front of things. I might be wrong but that was my impression of the current state of research 3 years ago.

 

[poetinsf]

The reason why the qualia of the sunset are relevant though is that they form a rich pattern that represents an outside scene and so it seems that something is being informed richly and in a way in which the richness provides detailed representational data.

But that richness is not conveyed on the signal. It is invoked and internally generated. Same thing happens in computer. An input to a function can trigger all kinds of cascading events and attach pre-stored data to it,