Tuesday, March 29, 2016

Human Consciousness, Artificial Intelligence and Popper’s Three World Ontology

Long-time commentator Ken B has drawn my attention to this post by Gene Callahan that complains that the Turing Test doesn’t really test for real intelligence.

In essence, I actually agree on this point, though I would reject Callahan’s dualism.

The endless debates about whether machines or software have real intelligence generally tend to suffer from a shoddy fallacy of equivocation. Generally, people who want to defend the real intelligence of machines make arguments like this:
(1) machines have intelligence and can think.

(2) a Turing test can show whether a machine has intelligence and can think.

(3) the brain is just an information processing machine, and so since computers are also information processing machines, a sufficiently complex computer should eventually have intelligence.
Yes, but what do you mean by “intelligence” and “thinking” in (1)?

Do you mean the conscious intelligence and thinking that I am experiencing right now? Actually, there is no good reason to think software can think in this way, because (3) is unproven and a non sequitur.

It does not follow that the consciousness of the human mind is just information processing that can be reproduced in other synthetic materials, such as in silicon chips in digital computers. After all, DNA and its behaviour exhibit a type of information processing and storage as well, but DNA is not conscious.

The conscious thinking intelligence that humans have is most probably a higher-level emergent property from specific biological processes in the brain.

As the analytic philosopher John Searle points out in the video below, all the empirical evidence suggests that consciousness is a biological phenomenon in the brain causally dependent on neuronal processes and biochemical activity, but one that can be explained by physicalist science, not some discredited supernatural ideas about souls or Cartesian dualism (see also Searle 1990; Searle 2002; Searle 1992).

Now the crude behaviourist Turing test does not test for whether an entity has conscious life, but merely whether it simulates human intelligence. Thus even if future computers all start passing Turing tests, it is not going to be some shocking milestone in human history: all it will show is that software programs have become sophisticated enough to fool us into thinking machines have conscious minds as we do, even though they do not.

Searle’s famous Chinese room argument is a devastating blow to the cult of AI.

But there are some additional philosophical points here. There is nothing supernatural about saying that, physically speaking, you need brain chemistry to produce a conscious intelligent mind.

And, philosophically speaking, the key to cracking the problem of human conscious intelligence is Karl Popper’s pluralist ontology. This is not a supernaturalist ontology, but naturalistic.

The key to understanding complexity in our universe is this: emergent properties. Karl Popper’s Three World Ontology gives us the best conceptual framework in which to understand our universe and the human mind.

Now this is where Searle goes philosophically wrong: his adherence to a crude monist materialism. Searle is unwilling to accept that (1) some emergent properties – like the human mind – are so remarkable that they make a new ontological category, and (2), because of (1), we have good reason to consider Popper’s three worlds ontology as a realistic metaphysics.

So what is Karl Popper’s pluralist ontology?

Karl Popper’s Three Worlds system is a pluralist ontology that classifies all objects in our universe, even though it is quite different from dualism. Ultimately, all objects in Popper’s world are causally dependent on matter and energy and their relations and permutations, so that – despite being ontologically pluralist – it has a fundamentally naturalist and physicalist/materialist basis.

We can set out the Three World Ontology in the following diagram.

World 1 give rises to World 2 objects, and World 2 in turn gives rise to World 3 objects, World 2 and World 3 being emergent properties from the each lower-level world respectively.

These three worlds can be described as follows:
World 1: the fundamental world of spatiotemporal physical objects and events, matter and energy and even biological systems considered merely as only complex physical systems. It further divided into a fundamental distinction between:
(1) the world of quantum mechanics, and

(2) the macroscopic world of Newtonian physics and Einstein’s special and general theories of relativity.
In some manner that is not properly understood our macroscopic world is an emergent property from the world of quantum mechanics.

World 2: the world of conscious intelligent human minds with intentionality and mental states (and, for example, any alien minds as complex as ours that have evolved on other planets, if they exist). True conscious intelligence is a property of minds in this world.

World 3: both
(1) unembodied objects such as abstract entities (and, most probably, universals), mathematical entities, social institutions and objective knowledge, and
(2) embodied entities that are products of human design, engineering and production such as buildings, books, clothes, money, and art works considered as human objects.
We must remember that “embodied entities” are members of both World 3 and World 1, but in different senses: e.g., considered purely physically as spatiotemporal objects describable in scientific terms, they belong to World 1; but when considered as human cultural artefacts, they belong to World 3.
The conscious intelligent human mind is an emergent property from World 1 brain states, but is so powerfully unique it requires a new ontological category.

A true artificial intelligence that would have the same type of conscious intelligent mind as the human mind will have to directly reproduce or replicate the biological processes in the brain that cause consciousness. An “artificial” intelligence – in the sense of not being a normal human being – will probably need to have organic or biochemical structures in its “brain” in order for it to be fully and truly conscious.

Such entities, if they were fully conscious, would create all sorts of ethical issues. They would probably have to be imbued with moral/ethical principles as humans are, for example. Probably they would have to be granted some kind of human rights at some point, so that we could not treat them as slaves. And what would they do? What work would they perform?

The whole notion that your desktop computer could become as conscious as you are should be a horrifying thought. Wouldn’t it effectively be a slave?

At any rate, the whole issue of whether truly conscious artificial intelligence can be created is a matter for a future science that has first completely mastered what human (and higher animal) consciousness actually is.

To return to the issue of ontology, Popper summed up his system as follows:
“To sum up, we arrive at the following picture of the universe.

There is the physical universe, world 1, with its most important sub-universe, that of the living organisms.

World 2, the world of conscious experience, emerges as an evolutionary product from the world of organisms.

World 3, the world of the products of the human mind, emerges as an evolutionary product from world 2.

In each of these cases, the emerging product has a tremendous feedback effect upon the world from which it emerged. For example, the physico-chemical composition of our atmosphere which contains so much oxygen is a product of life – a feedback effect of the life of plants. And, especially, the emergence of world 3 has a tremendous feedback effect upon world 2 and, through its intervention, upon world 1.

The feedback effect between world 3 and world 2 is of particular importance. Our minds are the creators of world 3; but world 3 in its turn not only informs our minds, but largely creates them. The very idea of a self depends on world 3 theories, especially upon a theory of time which underlies the identity of the self, the self of yesterday, of today, and of tomorrow. The learning of a language, which is a world 3 object, is itself partly a creative act and partly a feedback effect; and the full consciousness of self is anchored in our human language.

Our relationship to our work is a feedback relationship: our work grows through us, and we grow through our work.

This growth, this self-transcendence, has a rational side and a non-rational side. The creation of new ideas, of new theories, is partly non-rational. It is a matter of what is called ‘intuition’ or ‘imagination’. But intuition is fallible, as is everything human. Intuition must be controlled through rational criticism, which is the most important product of human language. This control through criticism is the rational aspect of the growth of knowledge and of our personal growth. It is one of the three most important things that make us human. The other two are compassion, and the consciousness of our fallibility.” (Popper 1978: 166–167).
At any rate, Popper’s three world ontology has the following virtues:
(1) it avoids the fallacy of strong reductionism;

(2) it is compatible with a moderate realist or conceptualist view of abstract entities and universals;

(3) it understands that human minds are an important, real entity in the universe and is consistent with John Searle’s biological naturalist theory of the mind, even though minds are a type of emergent physical property (for why Searle is not a property dualist, see Searle 2002);

(4) it is compatible with the finding of modern science that emergent properties are a fundamental phenomenon in our universe;

(5) it is compatible with downwards causation;

(6) it is ultimately consistent with the view that higher-level worlds are causally dependent for their existence on lower-level worlds.
Jesper Jespersen (2009) uses Popper’s three worlds ontology in his Critical Realist methodology for Post Keynesian economics, and it is highly relevant for the ontological and epistemological basis of economics.

Further Reading
“More on Karl Popper’s Three World Ontology,” September 6, 2013.

“Karl Popper’s Three World Ontology,” September 4, 2013.

“Limits of Artificial Intelligence,” February 24, 2014.

Jespersen, Jesper. 2009. Macroeconomic Methodology: A Post-Keynesian Perspective. Edward Elgar, Cheltenham, Glos. and Northampton, MA.

Popper, Karl R. 1978. “Three Worlds,” The Tanner Lecture on Human Values, Delivered at the University of Michigan, April 7, 1978.

Searle, John. R. 1980. “Minds, Brains, and Programs,” Behavioral and Brain Sciences 3: 417–424.

Searle, John. R. 1980. “Intrinsic Intentionality,” Behavioral and Brain Sciences 3: 450–456.

Searle, John R. 1982. “The Chinese Room Revisited,” The Behavioral and Brain Sciences 5: 345–348.

Searle, John. R. 1990. “Is the Brain a Digital Computer?,” Proceedings and Addresses of the American Philosophical Association 64: 21–37.

Searle, John R. 1992. The Rediscovery of the Mind. MIT Press, Cambridge, Mass and London.

Searle, J. 2002. “Why I Am Not a Property Dualist,” Journal of Consciousness Studies 9.12: 57–64.


  1. Have you seen this BBC documentary about AI and immortality?


    If you start 36 minutes into the documentary, a scientist describes his research in mapping the activity of a primitive biological brain (of the Hydra). If you fast forward to 48 minutes, he concludes that the type of activity observed that cannot be replicated by a computer.

    1. And he has published a mathematical proof, thus winning the Fields Medal?

      Confucius say when opponent wave arms like windmill you win.

    2. Oh, come on, Ken B.

      Don't you know the miserable history of failure of the old top-down "Good Old Fashioned AI" (GOFAI)/symbolic AI researchers when they thought they could create artificial intelligence as rich as human intelligence?

    3. Come on hell LK. Neurons have states that can be modeled. The claim advanced is that no computable function can describe the operation of these neurons. That would win the Fields and the biology Nobel, if it were true.

    4. You want to obfuscate by arguing every issue at once. Let's stick to this claim. Do you believe a finite number of neurons exhibit a mathematical function which cannot be calculated by any Turing machine? Do you believe this guy has a proof?

    5. Watch the BBC documentary "The Immortalist", Ken B:


    6. The latest research in neuroscience described towards the end of the documentary after the 36 minute mark vindicates Searle beautifully.

    7. "Do you believe a finite number of neurons exhibit a mathematical function which cannot be calculated by any Turing machine?"

      Ken B, you have misunderstand the issue here.

      Even if informational processing in Turing machines is a model of the input and output function of the human brain, and, yes, given the success of computers I think that is true, it still does not follow that our **conscious** intelligence per se can be reproduced inside digital computers and silicon.

      You can write a program that models digestion in the stomach, but can I feed food into my computer and het the biological process of digestion?
      Can you actually reproduce the physical process of digestion in software and digest food? No, of course you cannot.

      The analogy is very probably true of consciousness: even more, it is a higher level emergent property from very complex biological processes in the brain.

    8. That was not the claim made. A false claim was made, I refuted it. if I say you are wrong because you can buy a robot at Walmart that composes sonnets and symphonies, dances, sweats, and shits would it matter that this is a false claim? Then it matters that one of your useful idiots makes a false claim.

    9. Are you arguing brains digest food and excrete? Are you arguing that a man whose injury means he can no longer digest but must be fed by IV is not able to think? If you don't argue one of those how is your point about software eating pertinent? It's just another way you cite your meat prejudice as if it were evidence.

    10. Uggh, Ken B, [with me speaking in a calm and polite voice]:

      (1) "Are you arguing brains digest food and excrete? Are you arguing that a man whose injury means he can no longer digest but must be fed by IV is not able to think?"

      Of course not. I am saying that digestion is a process produced by **biological processes** in human stomachs. By analogy, human conscious intelligence is an also product of specific *biological* processes in the human brain (although obviously they are not the same biological processes, but both members of broad class of such various and different processes going on our body). Your desktop computer cannot digest food or experience true human consciousness because it lacks the necessary organic and biological structures to do so. Clear?

      (2) "If you don't argue one of those how is your point about software eating pertinent?"

      It is relevant as an analogy for what consciousness is: a product of biology.

      You ought at least to be receptive to the view that computers do not experience emotion, right? Obviously, because they do not have the necessary biology and hormone secretion? Similarly, consciousness is produced by specific biological processes. Computers lack this.

      (3) your point is that information processing goes on in computers? Correct. They are Turing machines. But information processing goes on in DNA too. DNA does not have the property of conscious intelligence.

    11. "You ought at least to be receptive to the view that computers do not experience emotion, right? "

      Since I am not discussing whether computers feel emotion I don't care. If you want to believe your old mum loves you because of a vital spark no machine can match I don't care. I am discussing if computers *think* and how we might judge.

      So let's recap. Callahan denied that a computer won at Go. He insisted that computers cannot think, and that the Turing test was stupid. He followed up with a series of posts showing his ignorance of the Turing test is so profound he cannot even get the mechanics right. Read Murphy and Tel and Barta on Murphy's, they are crushing. You brought shame on your family unto the seventh generation by agreeing with Callahan as he beclowned himself.

      So what I am discussing is 1) whether machines can think and 2) whether the Turing test is vacuous. I say yes and no; you two say no and yes. And all either of you have to support your contention is a prejudice about meat. "No matter what a machine can do, no matter how impossible it may be to tell it apart from meat, if it ain't meat it can't think." That is Callahan in a nutshell, and you seem to agree. So I will make a later, more provocative, comment on that when I get a chance.

    12. "So I will make a later, more provocative, comment on that when I get a chance."

      Oh, alright, then, Ken B, do your worst and take me to task for this heresy. lol.

    13. I was chatting with my neighbour, Lene Kallahan.

      Me: WTF. You deny the Holocaust?
      Lene: Of course.
      Me: WTF. You deny Hitler murdered 6 millions Jews??
      Lene: Of course. You can only murder human beings and Jews are not human, only simulacra.
      Me: WTF. "If you prick us, do we not bleed? If you tickle us, do we not laugh?"
      Lene: Oh the Shylock test. That's vacuous. If you fill a wineskin with blood and prick it, won't it bleed? That's worse than the Turing Test. Idiot.
      Me: WTF. Tell me then, how do you know that I am conscious?
      Lene: Oh I can tell that from talking to you. Obvioulsy you are a thinking being, just like all our neighbours, excpet Mr Cohen. I don't need a Turing test or a Shylock test; I just know.
      On the other hand, Jews may simulate thought and have bodily fluids but they have no consciousness or feelings.
      Me: WTF.On what possible basis can you assert that?
      Lene: Substrate. They are made of the wrong substrate.
      Me: WTF. You have evidence for this??
      Lene: My proof about computers just carries over, mutatis mutandis.
      Me: WTF.

      I walked away thinking Lene Kallahan is simply in the grip of a prejudice.

    14. Oh, christ, Ken B, what an atrocity of an argument..

      Since any group of human beings x would obviously have organic brains and brain processes (an MRI could easily establish this) that produce minds, one cannot make such an argument against my position.

      Human organic brains are clearly physically and chemically different from integrated circuits with electricity in computers.

    15. Kind of you to finger so precisely the prejudiced assumption.
      You haven't answered this. This is exactly the complaint Tel made against you. What you are saying is "I do not care what the Turing test shows or any test shows, it doesn't matter how often the criteria I use to know meat thinks also show silicon thinks, only meat thinks."

    16. Ok, Ken B, just tell me, how do you define "intelligence"?

      A clear definition, please.

    17. That's pretty much the point of the Turing test LK.
      Allow me to quote Turing's paper (for Gene's benefit; he hasn't read it):
      We may now consider again the point raised at the end of §3. It was suggested tentatively
      that the question, "Can machines think?" should be replaced by "Are there imaginable
      digital computers which would do well in the imitation game?".... The original question, "Can machines think?" I believe to be too
      meaningless to deserve discussion.

      The Turing Test is not, as Callahan pretends, a demand that we treat everyone we meet as a black box until we are satisfied. It is an attempt to *probe our own notions* and to *avoid extraneous assumptions which amount to prejudice*. This is perfectly explicit in the part I quoted.

      I think winning at Go shows thought and intelligence. I think following a scent does too, of a different kind. I even think being an Austrian economist does (YMMV).

      So I can no more define intelligence than you can. But if you propose a definition that includes the caveat "and on top of that only meat qualifies" I will object.

  2. Replies
    1. Thanks for the appreciation!

      After some reading I could also have said: the latest research in neuroscience vindicates Searle's views!

      See the work of Miguel Nicolelis (of the US Duke University), who argues that the conscious intelligent human mind just cannot be reproduced by a Turing Machine or digital computer, because the mind is produced by complex biological processes.

      See his book:

      Cicurel, Ronald and ‎Miguel A. L. Nicolelis. 2015. The Relativistic Brain: How it Works and Why it Cannot Be Simulated by a Turing Machine. Space Independent Publishing Platform.

      There is also a YouTube interview here, where he and a colleague refute the artificial intelligence cultists:


  3. "If you fast forward to 48 minutes, he concludes that the type of activity observed that cannot be replicated by a computer."

    Yep, that BBC documentary "The Immortalist" cites the work of Miguel Nicolelis (of Duke University), who, I think, rightly argues that you cannot reproduce the conscious human mind in in a Turing Machine or digital computer, because the brain is biological.

    Good discussion here:



    However, it is quite possible, as Nicolelis argues, that our biological brain will be augmented by artificial technology to enhance our thinking or perception.

    See also his book:

    Cicurel, Ronald and ‎Miguel A. L. Nicolelis. 2015. The Relativistic Brain: How it Works and Why it Cannot Be Simulated by a Turing Machine. Space Independent Publishing Platform.

  4. I don't think we can say whether a consciousness (World 2) is dependent on a specific platform or whether it is a general emergent property that can run on any computational platform. We cannot even say whether it emerges from the computation process, or is a property of some biological structure.

    Much of this uncertainty arises from the sad fact that we have no workable definition of consciousness, no way to measure it, and no theory to account for its workings.

  5. I expect we will see a gamut of intelligences, from 0 knowledge, 1 learning, 2 abstraction, 3 reflection/introspection, to 4 consciousness, 5 self consciousness, and many powerful systems that won't reach self consciousness. Dogs are but I wouldn't hope them slaves, but an intelligence greater than ours may see us as pets.

  6. Very interesting article, thanks for writing it.

    I think that consciousness likely being based on biology doesn't make it unachievable on computers, it just raises the bar for achieving it - eventually computers will be able to simulate the entire biological process within bodies, albeit this is an incredible level of complexity that would take hundreds, if not thousands, of years, to model exactly.

    Of course, it won't take exact modelling to be able to pull off 'consciousness' in AI - eventually the human body and mind will be understood well enough, to approximate this.

    Personally, I think that the whole idea of 'consciousness' is afforded an unwarranted level of 'specialness' - we don't yet understand what 'it' is or how it works exactly, but I think that the further the attempts to replicate the human brain go, the better our understanding of it will get.

  7. Have you read Roger Penrose's Emperor's New Mind and Shadows of the Mind. Shows mathematically that computers cannot possess intelligence like us humans.

    To show this, he constructs a question which cannot be answered by a computer. Even if the computer is fed with the answer, there's another question which the new computer cannot answer.

    1. Th book is based on a complete fallacy. (Mathematicans were stunned by Penrose's error.) This is not much disputed.

    2. I have not read this book but Landsburg recommends it highly. http://www.ams.org/notices/200703/rev-raatikainen.pdf
      (I did grad work on formal logic and all this stuff.)

  8. This comment has been removed by the author.

  9. If the claim is that biological nature of the brain is somehow necessary for consciousness, then one has to explain what it is about this biology that is so special that cannot be replicated in a computer. It is possible that this is the case but the burden of proof is on people claiming it.
    And Chinese room argument is in no way a "devastating blow" to anything. It shows a possible way for an entity passing a Turing Test not being conscious, but how does this necessarily mean that any such entity would not be conscious? And how does it show that consciousness is not an emergent phenomenon of a sufficiently complicated system even if it is on its basic level a Chinese room?
    For the record, I actually agree that there might be something special about the brain that straightforward AI attempts would not be able to replicate - and the only person I've seen who proposes anything at least testable and coherent is Scott Aaronson who suggests that there are reasons to doubt that consciousness can be cloned even in principle. If consciousness emerges somehow out of quantum states of the brain, then it is unclonable due to No-Cloning theorem. Aaronson lays out a bunch of paradoxes that arise when you consider the “naive” clonability view and establishes a possible alternative that can do away with these paradoxes:
    (and a much longer, but more detailed paper: http://www.scottaaronson.com/papers/giqtm3.pdf)