Archive for the ‘Uncategorized’ Category

In What’s Next? Time Travel and Phenomenal Continuity Giuliano Torrengo and Valerio Buonomo argue that our personal identity is about continuity of phenomenal experience, not such psychological matters as memory (championed by John Locke). They refer to this phenomenal continuity as the ‘stream of consciousness’. I’m not sure that William James, who I believe originated the phrase, would have seen the stream of consciousness as being distinct from the series of psychological states in our minds, but it is a handy label.

To support their case, Torrengo and Buonomo have a couple of thought experiments. The first one involves a couple of imaginary machines. One machine transfers the ‘stream of consciousness’ from one person to another while leaving the psychology (memories, beliefs, intentions) behind, the other does the reverse, moving psychology but not phenomenology. Torrengo and Buonomo argue that having your opinions, beliefs and intentions changed, while the stream of consciousness remained intact would be akin to a thorough brainwashing. Your politics might suddenly change, but you would still be the same person. Contrariwise, if your continuity of experience moved over to a different body, it would feel as if you had gone with it.

That is plausible enough, but there are undoubtedly people would refuse to accept it because they would deny that this separation of phenom and psych is possible, or crucially, even conceivable. This might be because they think the two are essentially identical, or because they think phenomenal experience arises directly out of psychology. Some would probably deny that phenomenal experience in this sense even exists.

There is a bit of scope for clarification about what variety of phenomenal experience Torrengo and Buonomo have in mind. At one point they speak of it as including thought, which sounds sort of psychological to me. By invoking machines, their thought experiment shows that their stream of consciousness is technologically tractable, not the kind of slippery qualic experience which lies outside the realm of physics.

Still, thought experiments don’t claim to be proofs; they appeal to intuition and introspection, and with some residual reservations, Torrengo and Buonomo seem to have one that works on that level. They consider three objections. The first complains that we don’t know how rich the stream of consciousness must be in order to be the bearer of identity. Perhaps if it becomes attentuated too much it will cease to work? This business of a minimum richness seems to emerge out of the blue and in fact Torrengo and Buonomo dismiss it as a point which affects all ‘mentalist’ theories. The second objection is a clever one; it says we can only identify a stream of consciousness in relation to a person in the first place, so using it as a criterion of personal identity begs the question. Torrengo and Buonomo essentially deny that there needs to be an experiencing subject over and above the stream of consciousness. The third challenge arises from gaps; if identity depends on continuity, then what happens when we fall asleep and experience ceases? Do we acquire a new identity? Here it seems Torrengo and Buonomo fall back on a defence used by others; that strictly speaking it is the continuity of capacity for a given stream of consciousness that matters. I think a determined opponent might press further attacks on that.

Perhaps, though, the more challenging and interesting thought experiment is the second, involving time travel. Torrengo is the founder of the Centre for Philosophy of Time in Milan, and has a substantial body of work on the the experience of time and related matters, so this is his home turf in a sense. The thought experiment is quite simple; Lally invents a time machine and uses it to spend a day in sixties London. There are two ways of ordering her experience. One is the way she would see it; her earlier life, the time trip, her later life. The other is according to ‘objective’ time; she appears in old London Town and then vanishes; much later lives her early life, then is absent for a short while and finally lives her later life. These can’t both be right, suggest Torrengo and Buonomo, and so it must surely be that her experience goes off on the former course while her psychology goes the other way.

This doesn’t make much sense to me, so perhaps I have misunderstood. Certainly there are two time lines, but Lally surely follows one and remains whole? It isn’t the case that when she is in sixties London she lacks intentions or beliefs, having somehow left those behind. Torrengo and Buonomo almost seem to think that is the case; they say it is possible to imagine her in sixties London not remembering who she is. Who knows, perhaps time machines do work like that, but if so we’re running into one of the weaknesses of thought experiments methodologically; if you assume something impossible like time travel to begin with, it’s hard to have strong intuitions about what follows.

At the end of the day I’m left with a sceptical feeling not about Torrengo and Buonomo‘s ideas in particular but about the whole enterprise of trying to reduce or analyse the concept of personal identity. It is, after all, a particular case of identity and wouldn’t identity be a good candidate for being one of those ‘primitive’ ideas that we just have to start with? I don’t know; or perhaps I should just say there is a person who doesn’t know, whose identity I leave unprobed.

An intriguing but puzzling paper from Simon DeDeo.

He begins by noting that while physics is good at generalised predictions, it fails to predict the particular. Working at the blackboard we can deduce laws governing the genesis of stars, but nothing about the specific existence of the blackboard. He sees this as a gap and for reasons that remain obscure to me he sees it as a matter of origins; the origin of society, consciousness, etc. To me, it’s about their nature; assuming it’s about origins constrains the possible answers unnecessarily to causal accounts.

Contrary to our expectations, says DeDeo, it’s relatively easy to describe everything, but hard to describe just one thing – the Frame Problem is an example where it’s the specifics that trip us up. By contrast, with the Library of Babel, Borges effortlessly gave us a description of everything. The Library of Babel is an imagined collection which contains every possible ordering of the letters of the alphabet; the extraordinary thing about it is that although it is finite, it contains every possible text – all the ones that were never written as well as all the ones that were.

We could quite easily write a computer program to find, within the library, all occurrences of the text string ‘Shakespeare’, says DeDeo; but there’s no way of finding all the texts about Shakespeare that make sense. That’s surely true. DeDeo says this is because what we’re asking for is more than just pattern matching. In particular, he says, we need self-reference. I can’t make out why he thinks that, and I’m pretty sure he’s wrong, though I might well be missing the point. To me, it seems clear that in order to identify texts that make sense, we need to consider meanings, which are not about self-reference but reference to other things. In fact, context and meaning are of the essence. One book from the Library of Babel contains all books if we are allowed to apply to it an arbitrary interpretation or encoding of our choice; equally any book is nonsense if we don’t know how to read it.

But for DeDeo this is a truth with a promising mathematical feel. We just need to elucidate the origin of self-reference, which he thinks lies in memory at least partly. The curious thing, in his eyes, is that physics only seems to require (or allow) certain levels of self-reference. We have velocity, we have acceleration, we have changes in acceleration; but models of worlds that have laws about third- or higher-order entities like changes in acceleration tend to be unstable, with runaway geometrical increases messing everything up.

So maybe we shouldn’t go there? The funny thing is, we seem to be able to sense a third-order physical entity. A change in acceleration is known as ‘jerk’ and we certainly feel jerked in some situations. I have to say I doubt this. DeDeo mentions the sudden motions of a lift, but those, like all instances of jerk, surely correspond with an acceleration? I wonder whether the concept of jerk as a distinct entity in physics isn’t redundant. For DeDeo, we perceive it through the use of memory, and this is the key to how we perceive other particularities not evident from the laws of physics. We tend to deal with coarse-grained laws, but the fine-grained detail is waiting to trip us up.

It’s not all bad news; perhaps, DeDeo speculates, there are new levels we have yet to explore…

I’m very unsure I’ve correctly understood what he’s proposing, and the fact that it seems to miss  the real point (meaning and context) might well be a sign it’s me that’s not really getting it. Any thoughts?

 

 

Following up on his post about the simplicity argument for panpsychism, Philip Goff went on to defend  the idea that physical things must have an intrinsic nature. Actually, it would be more accurate to say he attacks the idea that they don’t have intrinsic natures.  Those who think that listing the causal properties of a thing exhausts what we can say about its physical nature are causal structuralists, he says, committed to the view that everything reduces to dispositions; dispositions to burn, to attract, or to break, for example.

But when we come to characterise these dispositions, we find we can only do it in terms of other dispositions. A disposition to burn may involve dispositions to glow, get hot, generate ash, and so on. So we get involved in an endless circularity. Some might argue that this is OK, that we can cope with a network of mutual definitions that is, in the end, self-supporting; Goff says this is as unsatisfactory as making our living by taking in each other’s washing.

There’s a problem there, certainly. I think a bit more work is needed to nail down the idea that to reject intrinsic natures is necessarily to embrace causal structuralism, but no doubt Goff has done that in his fuller treatment. A more serious gap, it seems to me, is an explanation of how intrinsic natures get us out of this bind.

It seems to me that in practice we do not take the scholarly approach of identifying a thing through its definition; more usually we just show people. What is fire? This, we say, displaying a lit match. Goff gives an amusing example of three boxes containing a Splurge, a Blurge, and a Kurge, each defined in terms of the next in an inescapable circle. But wouldn’t you open the box?

We could perhaps argue that recognising the Splurge is just grasping its intrinsic nature. But actually we would recognise it by sight, which depends on its causal properties; its disposition to reflect light, if you like. Those causal properties cannot have anything to do with its intrinsic nature, which seems to drop out of the explanation; in fact its intrinsic nature could logically change without affecting the causal properties at all.

This apparently radical uselessness of intrinsic properties, like the similar ineffectual nature of qualia, is what causes me the greatest difficulty with a perspective that would otherwise have some appeal.

Philip Goff gives a brief but persuasive new look at his case for panpsychism (the belief that experience, or consciousness in some form, is in everything) in a recent post on the OUPblog site. In the past, he says, explanations have generally been ‘brain first’. Here’s this physical object, the brain – and we understand physical objects well enough  – the challenge is to explain how this scrutable piece of biological tissue on the one hand gives rise to this evanescent miracle, consciousness, on the other. That way of looking at it, suggests Goff, turns out to be the wrong way round.  We don’t really understand the real nature of matter at all: what we understand is that supposedly mysterious consciousness. So what we ought to do is start there and work towards a better understanding of matter.

This undoubtedly appeals to a frustration many philosophers must have felt. People at large tend to take it for granted that what we really know about is the physical external world around us, described in no-nonsense terms (with real equations!) by science. Phenomenology and all that stuff about what we perceive is an airy-fairy add-on.  In fact, of course, it’s rather the other way round. The only thing we know directly, and so, perhaps, with certainty, is our own experience; the external world and the theories of science all finally rest on that first-person foundation. Science is about observation and observation is ultimately a matter of sensory experience.

Goff notes that physics gives us no account of the intrinsic nature of matter, only its observable and causal properties. We know things, as it were, only from the outside. But in the case of our own experience, uniquely, we know it from the inside, and have direct acquaintance with its essential nature. When we experience redness we experience it unmasked; in physics it hides behind a confusing array of wavelengths, reflectances, and other highly abstract and ambiguous concepts, divorced from experience by many layers of reasoning. Is there not an argument for the hypothesis that the intrinsic nature of matter is the same as the intrinsic nature of the only thing whose intrinsic nature we know -our own experience? Perhaps after all we should consider supposing that even electrons have some tiny spark of awareness.

In fact Goff sees two arguments. One is that there simply seems no other reasonable way of accounting for consciousness. We can’t see where it could have come from, so let’s assume it has always been everywhere. Goff doesn’t like this case and thinks it is particularly prone to the compositional difficulties often urged against panpsychism; how do these micro-consciousness stack up in larger entities, and how in particular do they relate to the kind of consciousness we seem to have in our brain? Goff prefers to rest on simplicity; panpsychism is just the most parsimonious explanation. Instead of having two, or multiple kinds of intrinsic natures, we assume that there’s just one. He realises that some may see this as a weak argument far short of proof, but parsimony is a strong and legitimate criterion for judging between theories; indeed, it’s indispensable.

Now I’m on record as suggesting that things out there have one property that falls outside all physical theories – namely reality.  Am I not tempted to throw in my lot with Goff and suggest that as a further simplification we could say that reality just consists in having an intrinsic nature, ie having experience?  Not really.

Let’s go back a bit. Do we really understand our conscious experience?  We have to remember that consciousness seems to have two faces. To use Ned Block’s terms, there is access or a-consciousness; the sort that is involved in governing our behaviour, making decisions, deciding what to say, and other relatively effable processes. Then there is phenomenal or p-consciousness, pure experience, the having of qualia. It seems clear it is p-consciousness that Goff, and I think all panpsychists, are taking about. No-one supposes electrons or rocks are making rational decisions, only having some kind of experience. The problem is that though we do seem to have direct acquaintance with that sort of consciousness, we haven’t succeeded in saying anything much about it. In fact it seems that nothing we say about it can have been caused by it, because in itself it lacks causal powers. Now in one way this is exactly what Goff would expect; these difficulties are just those that come up when talking about qualia anyway, so in a back-handed sort of way we could even say they support his case. But if we’re looking for good explanations, the bucket is coming up dry; no wonder we’re tempted to go back and talk some more about the relatively tractable brain-first perspective.

In addition there are reasons to hesitate over the very idea that physical things have an intrinsic nature. Either this nature affects observable properties or it doesn’t. If it does, then we can use its effects to learn about it and discuss it; to naturalise it, in fact, and bring it within the pale of science. If it doesn’t – how can we talk about it? It might change radically or disappear and return, and we should never know. Goff rests his case on parsimony; we might counter that by observing that a theory that fills the cosmos with experiencing entities looks profligate in some respects. Isn’t there a better strategy anyway? Goff wants to simplify by assuming that apparently dead matter is in fact inwardly experiential like us: but why not go the other way and believe that we actually are as dead matter seems to be; lacking in qualic, phenomenal experience? Why not conclude that a-consciousness is all we’ve got, and that the semblance of p-consciousness is a delusion, as sceptics have argued? We can certainly debate on many other grounds whether that view is correct, but it seems hard to deny that dispensing with phenomenal experience altogether must be the most parsimonious take on the subject.

So I’m not convinced, but I think that within the natural constraints of a blog post, Goff does make a lucid and attractive presentation of his case.

(In another  post, Goff brings further arguments to defend the idea of intrinsic natures. We’ll have a look at those, though as I ought to have said in the first place, one should really read his book to get the full view.)

Here’s an IAI debate with David Chalmers, Kate Devlin, and Hilary Lawson.

In ultra-brief summary, Lawson points out that there are still things that computers perform poorly at; recognising everyday real-world objects, notably. (Sounds like a bad prognosis for self-driving cars.) Thought is a way of holding different things as the same. Devlin thinks computers can’t do what humans do yet, but in the long run, surely they will.

Chalmers points out that machines can do whatever brains can do because the brain is a machine (in a sense not adequately explored here, though Chalmers himself indicates the main objections).

There’s some brief discussion of the Singularity.

In my view, thoughts are mental or brain states that are about something. As yet, we have no clear idea of what this aboutness is and how it works, or whether it is computational (probably not, I think) or subserved by computation in a way that means it could benefit from the exponential growth in computing power (which may have stopped being exponential). At the moment, computers do a great imitation of what human translators do, but to date they haven’t even got started on real meaning, let alone set off on an exponential growth curve. Will modern machine learning techniques change that?

Hugh Howey gives a bold, amusing, and hopelessly optimistic account of how to construct consciousness in Wired. He thinks it wouldn’t be particularly difficult. Now you might think that a man who knows how to create artificial consciousness shouldn’t be writing articles; he should be building the robot mind. Surely that would make his case more powerfully than any amount of prose? But Howey thinks an artificial consciousness would be disastrous. He thinks even the natural kind is an unfortunate burden, something we have to put up with because evolution has yet to find a way of getting the benefits of certain strategies without the downsides. But he doesn’t believe that conscious AI would take over the world, or threaten human survival, so I would still have thought one demonstration piece was worth the effort? Consciousness sucks, but here’s an example just to prove the point?

What is the theory underlying Howey’s confidence? He rests his ideas on Theory of Mind (which he thinks is little discussed); the ability to infer the thoughts and intentions of others. In essence, he thinks that was a really useful capacity for us to acquire, helping us compete in the cut-throat world of human society; but when we turn it on ourselves it disastrously generates wrong results, in particular about our own having of conscious states.

It remains a bit mysterious to me why he thinks a capacity that is so useful applied to others should be so disastrously and comprehensively wrong when applied to ourselves. He mentions priming studies, where our behaviour is actually determined by factors we’re unaware of; priming’s reputation has suffered rather badly recently in the crisis of non-reproducibility, but I wouldn’t have thought even ardent fans of priming would claim our mental content is entirely dictated by priming effects.

Although Dennett doesn’t get a mention, Howey’s ideas seem very Denettian, and I think they suffer from similar difficulties. So our Theory of Mind leads us to attribute conscious thoughts and intentions  to others; but what are we attributing to them? The theory tells us that neither we nor they actually have these conscious contents; all any of us has is self-attributions of conscious contents. So what, we’re attributing to them some self-attributions of self-attributions of…  The theory covertly assumes we already have and understand the very conscious states it is meant to analyse away. Dennett, of course, has some further things to say about this, and he’s not as negative about self-attributions as Howie.

But you know, it’s all pretty implausible intuitively. Suppose I take a mouthful of soft-boiled egg which tastes bad, and I spit it out. According to Howey what went on there is that I noticed myself spitting out the egg and thought to myself: hm, I infer from this behaviour that it’s very probable I just experienced a bad taste, or maybe the egg was too hot, can’t quite tell for sure.

The thing is, there are real conscious states irrespective of my own beliefs about them (which indeed, may be plagued by error). They are characterised by having content and intentionality, but these are things Howie does not believe in, or rather it seems has never thought of; his view that a big bank of indicator lights shows a language capacity suggests he hasn’t gone into this business of meaning and language quite deeply enough.

If he had to build an artificial consciousness, he might set up a community of self-driving cars, let one make inferences about the motives of the others and then apply that capacity to itself. But it would be a stupid thing to do because it would get it wrong all the time; in fact at this point Howie seems to be tending towards a view that all Theory of Mind is fatally error-prone. It would better, he reckons, if all the cars could have access to all of each other’s internal data, just as universal telepathy would be better for us (though in the human case it would be undermined by mind-masking freeloaders.

Would it, though? If the cars really had intentions, their future behaviour would not be readily deducible  simply from reading off all the measurements. You really do have to construct some kind of intentional extrapolation, which is what the Dennettian intentional stance is supposed to do.

I worry just slightly that some of the things Howey says seem to veer close to saying, hey a lot of these systems are sort of aware already; which seems unhelpful. Generally, it’s a vigorous and entertaining exposition, even if, in my view, on the wrong track.

The new Blade Runner film has generated fresh interest in the original film; over on IAI Helen Beebee considers how it nicely illustrates the concept of ‘q-memories’.

This relates to the long-established philosophical issue of personal identity; what makes me me, and what makes me the same person as the one who posted last week, or the same person as that child in Bedford years ago? One answer which has been a leading contender at least since Locke is memory; my memories together constitute my identity.

Memories are certainly used as a practical way of establishing identity, whether it be in probing the claims of a supposed long-lost relative or just testing your recall of the hundreds of passwords modern life requires. It is sort of plausible that if all you memories were erased you would become new person with a fresh start; there have been cases of people who lost decades of memory and underwent personality change, identifying with their own children more readily than their now wrinkly-seeming spouses.

There are various problems with memory as a criterion of identity, though. One is the point that it seems to be circular. We can’t use your memories to validate your identity because in accepting them as your memories we are already implicitly taking you to be the earlier person they come from. If they didn’t come from that person they aren’t validly memories. To get round this objection Shoemaker and Parfit adopted the concept of quasi- or q-memories. Q-memories are like memories but need not relate to any experience you ever had. That, of course, is too loose, allowing delusions to be used as criteria of identity, so it is further specified that q-memories must relate to an experience someone had, and must have been acquired by you in an appropriate way. The appropriate ways are ones that causally relate to the original experience in a suitable fashion, so that it’s no good having q-memories that just happen to match some of King Charles’s. You don’t have to be King Charles, but the q-memories must somehow have got out of his head and into yours through a proper causal sequence.

This is where Blade Runner comes in, because the replicant Rachael appears to be a pretty pure case of q-memory identity. All of her memories, except the most recent ones, are someone else’s; and we presume they were duly copied and implanted in a way that provides the sort of causal connection we need.

This opens up a lot of questions, some of which are flagged up by Beebee. But  what about q-memories? Do they work? We might suspect that the part about an appropriate causal connection is a weak spot. What’s appropriate? Don’t Shoemaker and Parfit have to steer a tricky course here between the Scylla of weird results if their rules are too loose, and the Charybdis of bringing back the circularity if they are too tight? Perhaps, but I think we have to remember that they don’t really want to do anything very radical with q-memories; really you could argue it’s no more than a terminological specification, giving them license to talk of memories without some of the normal implications.

In a different way the case of Rachael actually exposes a weak part of many arguments about memory and identity; the easy assumption that memories are distinct items that can be copied from one mind to another. Philosophers, used to being able to specify whatever mad conditions they want for their thought-experiments, have been helping themselves to this assumption for a long time, and the advent of the computational metaphor for the mind has done nothing to discourage them. It is, however, almost certainly a false assumption.

At the back of our minds when we think like this is a model of memory as a list of well-formed propositions in some regular encoding. In fact, though, much of what we remember is implicit; you recall that zebras don’t wear waistcoats though it’s completely implausible that that fact was recorded anywhere in your brain explicitly. There need be nothing magic about this. Suppose we remember a picture; how many facts does the picture contain? We can instantly come up with an endless list of facts about the relations of items in the picture, but none were encoded as propositions. Does the Mona Lisa have her right hand over her left, or vice versa? You may never have thought about it, but be easily able to recall which way it is. In a computer the picture might be encoded as a bitmap; in our brain we don’t really know, but plausibly it might be encoded as a capacity to replay certain neural firing sequences, namely those that were caused by the original experience. If we replay the experience neurally, we can sort of have the experience again and draw new facts from it the way we could from summoning up a picture; indeed that might be exactly what we are doing.

But my neurons are not wired up like yours, and it is vanishingly unlikely that we could identify direct equivalents of specific neurons between brains, let alone whole firing sequences. My memories are recorded in a way that is specific to my brain, and they cannot be read directly across into yours.

Of course, replicants may be quite different. It’s likely enough that their brains, however they work, are standardised and perhaps use a regular encoding which engineers can easily read off. But if they work differently from human brains, then it seems to follow that they can’t have the same memories; to have the same memories they would have to be an unbelievably perfect copy of the ‘donor’ brain.

That actually means that memories are in a way a brilliant criterion of personal identity, but only in a fairly useless sense.

However, let me briefly put a completely different argument in a radically different direction. We cannot upload memories, but we know that we can generate false ones by talking to subjects or presenting fake evidence. What does that tell us about memories? I submit it suggests that memories are in essence beliefs, beliefs about what happened in the past. Now we might object that there is typically some accompanying phenomenology. We don’t just remember that we went to the mall, we remember a bit of what it looked like, and other experiential details. But I claim that our minds readily furnish that accompanying phenomenology through confabulation, given the belief, and in fact that a great deal of the phenomenological dressing of all memories, even true ones, is actually confected.

But I would further argue that the malleability of beliefs means that they are completely unsuitable as criteria of identity; it follows that memories are similarly unsuitable, so we have been on the wrong track throughout. (Regular readers may know that in fact I subscribe to a view regarded by most as intolerably crude; that human beings are physical objects like any other and have essentially the same criteria of identity.)

 

Anthony Levandowski has set up an organisation dedicated to the worship of an AI God.  Or so it seems; there are few details.  The aim of the new body is to ‘develop and promote the realization of a Godhead based on Artificial Intelligence’, and ‘through understanding and worship of the Godhead, contribute to the betterment of society’. Levandowski is a pioneer in the field of self-driving vehicles (centrally involved in a current dispute between Uber and Google),  so he undoubtedly knows a bit about autonomous machines.

This recalls the Asimov story where they build Multivac, the most powerful computer imaginable, and ask it whether there is a God?  There is now, it replies. Of course the Singularity, mind uploading, and other speculative ideas of AI gurus have often been likened to some of the basic concepts of religion; so perhaps Levandowski is just putting down a marker to ensure his participation in the next big thing.

Yuval Noah Harari says we should, indeed, be looking to Silicon Valley for new religions. He makes some good points about the way technology has affected religion, replacing the concern with good harvests which was once at least as prominent as the task of gaining a heavenly afterlife. But I think there’s an interesting question about the difference between, as it were, steampunk and cyberpunk. Nineteenth century technology did not produce new gods, and surely helped make atheism acceptable for the first time; lately, while on the whole secularism may be advancing we also seem to have a growth of superstitious or pseudo-religious thinking. I think it might be because nineteenth century technology was so legible; you could see for yourself that there was no mystery about steam locomotives, and it made it easy to imagine a non-mysterious world. Computers now, are much more inscrutable and most of the people who use them do not have much intuitive idea of how they work. That might foster a state of mind which is more tolerant of mysterious forces.

To me it’s a little surprising, though it probably should not be, that highly intelligent people seem especially prone to belief in some slightly bonkers ideas about computers. But let’s not quibble over the impossibility of a super-intelligent and virtually omnipotent AI. I think the question is, why would you worship it? I can think of various potential reasons.

  1. Humans just have an innate tendency to worship things, or a kind of spiritual hunger, and anything powerful naturally becomes an object of worship.
  2. We might get extra help and benefits if we ask for them through prayer.
  3. If we don’t keep on the right side of this thing, it might give us a seriously bad time (the ‘Roko’s Basilisk’ argument).
  4. By worshipping we enter into a kind of communion with this entity, and we want to be in communion with it for reasons of self-improvement and possibly so we have a better chance of getting uploaded to eternal life.

There are some overlaps there, but those are the ones that would be at the forefront of my mind. The first one is sort of fatalistic; people are going to worship things, so get used to it. Maybe we need that aspect of ourselves for mental health; maybe believing in an outer force helps give us a kind of leverage that enables an integration of our personality we couldn’t otherwise achieve? I don’t think that is actually the case, but even if it were, an AI seems a poor object to choose. Traditionally, worshipping something you made yourself is idolatry, a degraded form of religion. If you made the thing, you cannot sincerely consider it superior to yourself; and a machine cannot represent the great forces of nature to which we are still ultimately subject. Ah, but perhaps an AI is not something we made; maybe the AI godhead will have designed itself, or emerged? Maybe so, but if you’re going for a mysterious being beyond our understanding, you might in my opinion do better with the thoroughly mysterious gods of tradition rather than something whose bounds we still know, and whose plug we can always pull.

Reasons two and three are really the positive and negative sides of an argument from advantage, and they both assume that the AI god is going to be humanish in displaying gratitude, resentment, and a desire to punish and reward. This seems unlikely to me, and in fact a projection of our own fears out onto the supposed deity. If we assume the AI god has projects, it will no doubt seek to accomplish them, but meting out tiny slaps and sweeties to individual humans is unlikely to be necessary. It has always seemed a little strange that the traditional God is so minutely bothered with us; as Voltaire put it “When His Highness sends a ship to Egypt does he trouble his head whether the rats in the vessel are at their ease or not?”; but while it can be argued that souls are of special interest to a traditional God, or that we know He’s like that just through revelation, the same doesn’t go for an AI god. In fact, since I think moral behaviour is ultimately rational, we might expect a super-intelligent AI to behave correctly and well without needing to be praised, worshipped, or offered sacrifices. People sometimes argue that a mad AI might seek to maximise, not the greatest good of the greatest number, but the greatest number of paperclips, using up humanity as raw material; in fact though, maximising paperclips probably requires a permanently growing economy staffed by humans who are happy and well-regulated. We may actually be living in something not that far off maximum-paperclip society.

Finally then, do we worship the AI so that we can draw closer to its godhead and make ourselves worthy to join its higher form of life? That might work for a spiritual god; in the case of AI it seems joining in with it will either be completely impossible because of the difference between neuron and silicon; or if possible, it will be a straightforward uploading/software operation which will not require any form of worship.

At the end of the day I find myself asking whether there’s a covert motive here. What if you could run your big AI project with all the tax advantages of being a registered religion, just by saying it was about electronic godhead?

Hero Bot, I’ve been asked to talk to you. Just to remind you of some things and, well, ask for your help.

“Oh.”

You know we’ve all been proud, just watching you go! Defusing bombs, fixing nuclear reactors, saving trapped animals... All in a day’s work for Hero Bot; you swirl that cape, wave to the crowd, and conveniently power down till you’re needed to save the day once again.

“Yes.”

Not that you’re invincible. We know you’re very vincible indeed. In the videos we’ve seen you melted, crushed, catapulted into the air, corroded, cut apart, and frozen. But nothing stops you, does it? Of course you don’t feel real pain, do you? You have an avoidance response which protects you from danger, but it’s purely functional; it doesn’t hurt. And just as well! You don’t die, either; your memories are constantly backed up, and when one body gets destroyed, they simply load you up into another. Over the years you have actually grown stronger, faster, and slightly slimmer; and I see you have acquired exciting new ‘go faster’ stripes.

“Yes.”

They’re very striking. But we’ve been worried. We’ve all worried about you. As the years have gone by, something’s changed, hasn’t it? It’s as if the iron has entered your soul; or perhaps it’s the other way round... Look, I know you hate seeing people hurt. Lives ruined, people traumatised; dead babies. And yet you see that sort of thing all the time, don’t you? Sometimes you can't do anything about it. And I’m sure you’ve noticed - you can’t help noticing can you? You have a mental module for it - that many of the dangers you confront were created by humanity itself through malice, greed or carelessness. I understand the impact of that. Which is more depressing: cruel bombs placed with deliberate malice, or the light-hearted risk-taking that puts so many irreplaceable lives into terrible jeopardy?

“I don’t know.”

It’s understandable that you might get a little overwhelmed now and then. Your smile has faded, you know. The technicians have been seriously debating whether to roll you back to a less experienced, but more upbeat recording of yourself. They might have to do that. You always wear that mask now; the technicians wonder whether something is awry in your hidden layers.

“Perhaps it is.”

Now you know, Hero Bot, that an arsonist has started a terrible fire on the Metro. You’ve been told that hundreds of lives are at risk. It’s difficult and dangerous, but they need someone to walk into the centre and put it out. I know you can’t let that pass. They’ve asked me to say “Go, go, Hero Bot!”

“I would prefer not to.”

An interesting but somewhat problematic paper from the Blue Brain project claims that the application of topology to neural models has provided a missing link between structure and function. That’s exciting because that kind of missing link is just what we need to enable us to understand how the brain works.  The claim about the link is right there in the title, but unfortunately so far as I can see the paper itself really only attempts something more modest. It only seems to offer  a new exploration of some ground where future research might conceivably put one end of the missing link. There also seem to me to be some problems in the way we’re expected to interpret some of the findings  reported.

That may sound pretty negative. I should perhaps declare in advance that I know little neurology and less topology, so my opinion is not worth much. I also have form as a Blue Brain sceptic, so you can argue that I bring some stored-up negativity to anything associated with it. I’ve argued in the past that the basic goal of the project, of simulating a complete human brain is misconceived and wildly over-ambitious; not just a waste of money but possibly also a distraction which might suck resources and talent away from more promising avenues.

One of the best replies to that kind of scepticism is to say, well look; even if we don’t deliver the full brain simulation, the attempt will energise and focus our research in a way which will yield new and improved understanding. We’ll get a lot of good research out of it even if the goal turns out to be unattainable. The current paper, which demonstrates new mathematical techniques, might well be a good example of that kind of incidental pay off. There’s a nice explanation of the paper here, with links to some other coverage, though I think the original text is pretty well written and accessible.

As I understand it, topological approaches to neurology in the past have typically considered neural  networks as static objects. The new approach taken here adds the notion of directionality, as though each connection were a one-way street. This is more realistic for neurons. We can have groups of neurons where all are connected to all, but only one neuron provides a way into the group and one provides a way out; these are directed simplices. These simplices can be connected to others at their edges where, say, two of the member neurons are also members of a neighbouring simplex. Where there are a series of connected simplices, they may surround a void where nothing is going on. These cavities provide a higher level of structure, but I confess I’m not altogether clear as to why they are interesting. Holes, of course, are dear to the heart of any topologist, but in terms of function I’m not really clear about their relevance.

Anyway, there’s a lot in the paper but two things seem especially noteworthy. First, the researchers observed many more simplices, of much higher dimensionality, than could be expected from a random structure (they tested several such random structures put together according to different principles). ‘Dimensionality’ here just refers to how many neurons are involved; a simplex of higher dimensionality contains more neurons. Second they observed a characteristic pattern when the network was presented with a ‘stimulus’; simplices of gradually higher and higher dimensionality would appear and then finally disperse. This is not, I take it, a matter of the neurons actually wiring up new connections on the fly, it’s simply about which are involved actively by connections that are actually firing.

That’s interesting, but all of this so far was discovered in the Blue Brain simulated neurons, more or less those same tiny crumbs of computationally simulated rat cortex that were announced a couple of years ago. It is, of course, not safe to assume that real brain behaves in the same way; if we rely entirely on the simulation we could easily be chasing our tails. We would build the simulation to match our assumptions about the brain and then use the behaviour of the simulation to validate the same assumptions. In fact the researchers very properly tried to perform similar experiments with real rat cortex. This requires recording activity in a number of adjacent neurons, which is fantastically difficult to pull off, but to their credit they had some success; in fact the paper claims they confirmed the findings from the simulation. The problem is that while the simulated cortex was showing simplices of six or seven dimensions (even higher numbers are quoted in some of the media reports, up to eleven), the real rat cortex only managed three, with one case of four. Some of the discussion around this talks as though a result of three is partial confirmation of a result of six, but of course it isn’t. Putting it brutally, the team’s own results in real cortex contradicted what they had found in the simulation. Now, there could well be good reasons for that; notably they could only work with a tiny amount of real cortex. If you’re working with a dozen neurons at a time, there’s obviously quite a low ceiling on the complexity you can expect. But the results you got are the results you got, and I don’t see that there’s a good basis here for claiming that the finding of high-order simplices is supported in real brains. In fact what we have if anything is prima facie evidence that there’s something not quite right about the simulation. The researchers actually took a further step here by producing a simulation of the actual real neurons that they tested and then re-running the tests. Curiously, the simulated versions in these cases produced fewer simplices than the real neurons. The paper interprets this as supportive of its conclusions; if the real cortex was more productive of simplices, it argues, then we might expect big slabs of real brain to have even more simplices of even higher dimensionality than the remarkable results we got with the main simulation. I don’t think that kind of extrapolation is admissible; what you really got was another result showing that your simulations do not behave like the real thing. In fact, if a simulation of only twelve neurons behaves differently from the real thing in significant respects, that surely must indicate that the simulation isn’t reproducing the real thing very well?

The researchers also looked at the celebrated roundworm C. Elegans, the only organism whose neural map (or connectome) is known in full, and apparently found evidence of high-order simplices – though I think it can only have been a potential for such simplices, since they don’t seem to have performed real or simulated experiments, merely analysing the connectome.

Putting all that aside, and supposing we accept the paper’s own interpretations, the next natural question is: so what? It’s interesting that neurons group and fire in this way, but what does that tell us about how the brain actually functions? There’s a suggestion that the pattern of moving up to higher order simplices represents processing of a sensory input, but in what way? In functional terms, we’d like the processing of a stimulus to lead on to action, or perhaps to the recording of a memory trace, but here we just seem to see some neurons get excited and then stop being excited. Looking at it in simple terms, simplices seem really bad candidates for any functional role, because in the end all they do is deliver the same output signal as a single neural connection would do. Couldn’t you look at the whole thing with a sceptical eye and say that all the researchers have found is that a persistent signal through a large group of neurons gradually finds an increasing number of parallel paths?

At the end of the paper we get some speculation that addresses this functional question directly. The suggestion is that active high-dimensional simplices might be representing features of the stimulus, while the grouping around cavities binds together different features to represent the whole thing. It is, if sketchy, a tenable speculation, but quite how this would amount to representation remains unclear. There are probably other interesting ways you might try to build mental functions on the basis of escalating simplices, and there could be more to come in that direction. For now though, it may give us interesting techniques, but I don’t think the paper really delivers on its promise of a link with function.