Some thalamic-cortical connections are widely (and historically) well known. Like its input to layer 4, while other cortico-cortical connections are not so widely examined (there are few trace studies here and there). One of the tenets of TBT, is attachment of reference frames to each sensory input (as well as up into any hierarchy.
Now into my intuitive proof. It has to do with Qualia. Just a little explanation to direct everyone to the same page. When people talk about consciousness they mean a lot of different things. Generally, self-awareness, subjective-experience and qualia are some of these things.
Letās further break down the concept of qualia. Qualia itself, can be broken down to all the different sensory feelings. The redness of red, painfulness of pain, emotions, and so on ā¦
Of these different qualias, one caught my eye a few month ago. EVERY one is having an out of the body experience every moment of their lives. Close your eyes, and touch something (closing your eyes is not necessary but will help), you will feel something at the tip of you finger.
Thatās SO weird. your brain is generating a learned map of your touch outside your body ( you feel the touch at exact point that your body has learned). You can close your eyes, and touch something, you can still be damn sure of exact spatial (vector) of the point of your touch, as if your brain is attaching a reference Frame (which must have been learned to your sensory input (not necessarily learned)). Now, if thatās true for a finger touch, then it must be true for other sensory inputs, such as vision (or maybe its attached at perception level).
Now we all know about phantom hand, and it takes a long time to learn new reference frames.
Thought to share this as it really does feel like an intuitive proof of reference frames attachment to any sensory input.
That would be introspection. I could do that, or I could just tell you I had and you would have no way of telling which. Likewise qualia: do they exist? How do you know?
[Yes, I know that was probably tongue in cheek, but seriously ā does this stuff have any place here?]
Looking at all of the many speculations (some borne of introspection ) regarding the nature of human consciousness, most are agreed that it involves an internal mindspace as well as a timeline. What better to model it than HTM?
Perhaps this is amenable to verification via one of the many viewing tools that allow the state of the cortex to be viewed during tasks.
Subject A states that they are experiencing something and the tool shows some activation pattern. Subject B states they are experiencing the same thing and the same pattern is scanned.
I can imagine building a library of correlates where you can reliably predict what a subject would state they are experiencing if you were to ask.
Itās not purely the introspection as guidance but rather as a new evidence (in bayesian sense) to update the belief we have about the current parameters of our model (which is TBT).
Bayesian theorem, is in fact how ALL science is advanced. You have a hypothesis (or a model of how a phenomenon works), this hypothesis can be simple (like figuring out speed and momentum) or multidimensional (depending on the number of variables of your model such as all the parameters of a neural network), you obtain a new evidence and update your belief using the bayesian framework that you have. This could be just changing your probability distribution a tiny bit (depending on the weight of the new evidence).
Back to my theory, itās not something extraordinary crazy (I called it qualia, as itās what itās called) but you can call it whatever you like. Your sensual experience has a locality. The sensory feeling of your touch is not POSSIBLE to be felt without a location attached to it. You are simply incapable of that.
Thatās an evidence. How can we update the model we have about our brain using this tiny new little data we have? well, thatās what I proposed, that maybe all sensory inputs are infused with location data right from the get go.
Thatās SCIENCE. most hypothesis in science are initially formed using EDUCATED intuition, further after the birth of the hypothesis, itās falsified using scientific method.
Just because qualia gets a bad rep, it doesnāt mean that it canāt be a dim light in an extremely computationally complex circuitry of the brain.
Here is a thing, a lot of the scientific theories we take for granted are tautological. Theyāre inseparable from the suff that forms the theory itself. Take Darwinās survival of the fittest. Itās tautological as itās a theory of how survival works, by looking at things that have survived. All Iām saying is that sensory location data is SO infused in your touch, thatās it tautological. You have to know that, the afferent neurone that take the sensory input to your brain always send the same information regardless of your fingers vector location. Putting it other way, youāre sensual experience of your touch, is permanent. But the location of your finger changes all the time. Now if youāre incapable of feeling your touch, without knowing its location, somehow along the line, some location data MUST get attached to this sensory input. Iāll leave this post at that. Again itās tautological inference so itās a bit hard to get what I explained.
āMost are agreedā is not scienceā¦ How do you falsify speculation?
As I see it HTM is a potent model for certain kinds of pattern recognition, especially sequences and anomalies. It operates at the level of cortical columns, perhaps hundreds of neurons but not millions.
Consciousness is experienced as a curious mix of attention, language, story-telling, visualisation and explication, relying heavily on lower level brain function to which we have no direct conscious access. That would place it at a much higher level than HTM.
So my question, as before, is how to make that transition, from an intuition possibly guided by introspection to actual science, in which introspection has no role. What is your hypothesis? What prediction can you make? What empirical data can you gather?
Well, in this particular direction of research, as long as one cannot prove how a set of certain theories are essential to and can be combined up into a functional system implementing an intelligent agent, then all these theories are in the waiting queue for falsification.
Meh, thatās what I doubt mostly. According to my theory, we first build a map of the body and connecting touch sensations with regions on the map is both learned and speculated** the hard way.
** By speculation I mean a simplifying assumption about present evidence. Contrary to some views, it is essential to human learning, cognition and ability to build a unified, imperfect yet useful and improvable model of reality with very limited amounts of data (aka short life experience)
No matter, I concluded four years ago that the only way to prove my own conclusions regarding machine sentience is to just build one. HTM is not (my opinion here) the neural correlate for consciousness, but the underlying architecture on which it is built.
Introspection is not science, it is a source of ideas which can be turned into hypothesis which can be tested against simulations and or biology. Aristotleās atomistic idea was not science but could be tested with the right instruments/tests.
Not really. Aristotleās ideas were wrong, but they did provide some impetus for a process to prove them wrong, which in some ways led to science. You might be thinking of Democritus (who coined the term and the concept), but the real work starts with Dalton.
And introspection has nothing in common with ideas. Introspection is the study of oneās own thoughts, feelings and mental processes. It has its uses in self-discovery, not not in science. You can have ideas about anything, but you can only introspect on your own mind and how you think it works (which is probably just as wrong as Aristotle).
Iād argue that every single novel idea/innovation ever conceived by a human mind is by definition the product of introspection, since the person had to dive on its own thoughts and combine them in a way that has never been done before, or at least never exposed to the outside world before.
Also, I have the impression that this discussion wonāt go anywhere if we dont come up with a very clear and āsharpā definition for introspection. The idea I have about introspection is too fuzzy and abstract to work with.
āStudy of ones own mental processesā wonāt do it for me, it literally encompasses what I did every second of my life. Is it even possible to think without introspecting? I feel like the only way I could possibly do that is by surgically removing my frontal lobe. See what Iām doing right now, isnāt that introspecting?
I think you have to agree that philosophy has a large introspection component. Descartes was king with cogito ergo sum, and then Dennett came a long and introspected the āCartesian Theatreā and qualia away. Then Dennett went on to develop, now with science, his Multiple Drafts theory. Take a hard look at āThousand Brainsā and show me how it isnāt the same thing. Nod to Hawkins for taking it to science, philosophy mucks things up with all the introspection, but they do get the ball rolling.
The OP is way too domain-specific to TBT and the concept of reference frames, their specific relationship with sensor location, and the idea of them being universally applied to all sensors. GPT-3 is good, but Iāve yet to see it quite that on-point with technical details when primed with HTM-related inputs (which I have done a number of times out of curiosity about what it might spit out).
Now human assisted/curated GPT-3 output, definitely possible. However, in this case, I really think @walkertalker was trying to convey a legitimate āahaā moment (similar to the one Jeff describes in his book). The fact that one can āsenseā the location of an input along with its raw data, is itself evidence that the brain must have some mechanism for encoding locations.
Now the question of whether that evidence is useful in the scientific process to support or falsify any hypothesis, is of course an important point. I personally think it can be used as (perhaps weak) supporting evidence for a hypothesis. It would probably be pretty challenging to use this type of evidence for falsifying a hypothesis, though (which of course makes it less useful than many other types of evidence).
) regarding the nature of human consciousness, most are agreed that it involves an internal mindspace as well as a timeline. What better to model it than HTM?