@T: ALWAYS THE TWAIN

Maybe I could have entitled it: “Twaining a Conscious Web”

I decided to write this post because of the passionate debates a handful of us have been having about AI, NLP and resolving the reducibility question. At the surface, this may seem strange or over theoretical but at its core it goes into how much more powerful and intelligent the machines we build can be and also whether there is such a thing as data consciousness, context and the ability to compute every theory.

If any readers think that 50-100 comments on a thread is substantial (and whenever I’m on a thread, they do seem to attract more comments from others — it’s quite bizarre how interested people become and want to actively contribute which is good), this seriously smart debate is a 530+ strong comments thread!!! So…………..at some point we’ll be able to solve the question of how the Universe was created — LOL.

Essentially, there are two key players whose work is being debated:

* Gregory Chaitin — data is not reducible and there are some things we just can’t know.

* Stephen Wolfram (of Wolfram Alpha) — data is reducible and we can compute the answer to every question.

Computer scientists are currently trying to break data down into forms and associations that can be more readily computed, connected and extracted. It’s well known that in NLP, meaning extraction is still proving to be problematic. Additionally, we have to factor in Nicholas Taleb’s “Black Swan” type probability anomalies if we’re going to be able to compute everything, apparently.

[Twain's observation: we can't currently compute the answer to every question --- for example, how much and why we love our parents, the existence of a Supreme Being or anything involving subjectivity that's culture affiliated --- and what we should be doing is innovating algorithms to contextualize as much of the data points as possible. Moreover, data is reducible and can be transformed for smarter extraction in ways we haven't explored yet.]

As well as the Chaitin and Wolfram, we also have to throw Godel’s two Incompleteness Theorems — which work their way through mathematical logic like so:

After all that (con)textual mathematics, here are the key Chaitin videos which encapsulate his position on “maximum unknowns”.

Now here’s Stephen Wolfram explaining the computability of everything:

So then how does social science and psychology strand into what is high-end mathematics? Well……….

Those of us who’ve studied macroeconomics are aware of the J curve theory from John Maynard Keynes:

A country’s trade deficit will worsen initially after the depreciation of its currency because higher prices on foreign imports will be greater than the reduced volume of imports.

The J curve theory has been adapted by management consultants like Gartner into a theory about technology hype cycles:

Meanwhile, Otto Scharmer in organizational behavior has proposed a different letter from J, U, to explain how we examine ourselves, our perspectives on the world and the way in which we solve problems:

Since both camps (vectorial scale algorithms approach to data reducibility versus Quantum Mechanics approach) have merits, I wrote:

This thread is observing a classic W approach to problem-solving: two schools of thought, approaching from either end, drilling down and shifting their vectorial positions as time elapses and finally inflecting upwards (with views on what their prior slopes looked like) until the two schools converge and are on a different plane from where each and both started.

Wrt whether the Semantic Web can become an inference engine………….not if it continues to deploy the taxonomies and categorizations it does because that still roots us in probabilities, correlations and the other facets of it being actually no fundamentally different from Google (which itself is a difference engine just as Babbage postulated — albeit instead of absolute real numbers and binaries, it’s about the difference between statistical points).

Now, if it was a …..true DIFFERENTIATION ENGINE, this would be a real leap forward rather than an imaginary one.

We cannot infer until we can differentiate and the Semantic Web cannot do this (yet).

Specifically on data reducibility and context, this is what I offered to the debate:

The dimensions of context for each of us is personal, experiential, spiritual / emotive and cultural.

Conversely, the dimensions of computing are impersonal, iterative, rational and culture agnostic.

Explications which seem clear, obvious and even underpinned by established/irrefutable science and mathematical equations are fine to follow if we’re conversing with another person schooled with the same scientific reference points as us. They’re not so clear, obvious and irrefutable when we’re conversing with a lawyer, a photographer or a linguist because their context points for deduction, perspective and language will be distributively different.

This is why when we surface a piece of raw data in a search engine list – suppose something as simple as the number 2 – the context of it is going to be interpreted in diverse ways. The literally-minded will perceive it just as the number after 1 with a value of 2. The mathematically-minded will think of it as being a prime as well as in terms of power series, halves and double integrals. The literately-minded will automatically associate it with ‘Tale of Two Cities’, “To be or not to be”, JRR Tolkien (‘Two Towers’) and “it takes two to tango”. The artistically-minded will see the image of a swan or one half of a heart because that’s the shape of a 2. The Spanish-speaking computer scientist will think of it as being DOS (Disk Operating System). The classical scholar will reference it to Janus, the god of two faces. The romantic would tie it with coupledom…..etcetcetc………Whilst the Chinese would word associate it with the homophone for “fish”, “happiness” and “prosperity” all intrinsically bound to each other.

So that’s an example of raw data carrying implicit context which is not currently being included in or extracted explicitly by algorithms.

How to resolve this so we can compute this context and not only the raw data?

I’m working on it, as they say.

And so……………..I am………………

Yesterday in the Times there was an article entitled, ‘Babies’ brains are more sophisticated than we ever believed’:

* http://women.timesonline.co.uk/tol/life_and_style/women/families/article6793658.ece

Quite a few of the readers’ comments on the Times’s article are interesting, revealing and worth reading.

TWAIN’S VIEWS

Well, I’ve known how much brighter babies are than adults give them credit for ever since I was a baby myself, later when my younger brother was born and now when I see babies out and about. Yesterday, the whole of London seemed to bloom with babies since the sunshine brought out all their proud parents and their prams. Some were also in harnesses on their mothers’ backs or making their first attempts at walking. One little baby girl in a bright pink baby suit decided she couldn’t figure out which order her feet were supposed to go to make a step, so she plunged herself onto her derrière in protest — LOL. She’ll probably grow up to be a campaigner and do sit-ins. Another baby decided he wanted to show off his ability to put his big toes into his mouth. Cute.

Anyway, I didn’t need any Barbies / Sindys / teddy bears as a kid because I had a real live, kicking, screaming, gurgling, learning human (Twain) experiment, bundle of joy in our family in the shape of our youngest. There was enough of an age gap between us for me to actually treat him like a scientific study case rather than just go “Goo-goo-ga-ga, awwwwww,” over him!

I used to put him through his paces to test his mental, physical, audio-visual and emotional dexterity and consciousness. That started happening when he was 3 months old and my parents decided they could trust us to help him do his muscle strengthening exercises. This involved putting him on his back and gently stretching out his limbs, whilst counting to him in Mandarin with each movement and then holding up brightly colored objects to see whether he was:

(1.) able to follow the object around;

(2.) able to detect when an object had been swapped; and

(3.) able to anticipate whether we were going to put the object near his nose / his hands / us.

He was pretty good at all of these tests. His special talent was more audio than visual, though. Once he could walk the first thing he tried to do was switch the TV on for the sound. Maybe that explains his musical talents now.

Before he arrived though, I’d been experimental with kids younger than me when I was about 4. I babysat a neighbor’s little 2-year-old and earned HK$2 per day for my efforts. That baby, though, was definitely not as bright or inquisitive as my younger brother. She was quieter and more introspective.

With my own children — the ones that will make my mother a grandmother — I plan to record and document all my experiments with them. That’ll be fun!

T-MODEL OF CONSCIOUSNESS

So I was thinking about this whole issue of, “What’s consciousness and where is it,” my father’s coma situation and was also wondering whether the Internet’s version of the Global Brain might enable us to produce a proxy for our natural brain, the location of consciousness and this is what’s emerging as my model:

By medical definition, my father was considered to be “unconscious” which meant that the ECG (electrical conductivity graphs) couldn’t detect any discernible voltage that might indicate electrical activity in his brain. He also seemed to be unresponsive to instructions and actions from the nursing staff. Yet when I visited I got the distinctive impression that he was conscious so I set about doing my own experiments to test for his responsiveness — over and above whatever the ECGs and daily physical routines / procedures the hospital staff were doing.

I reasoned that, according to medical literature, we’ve identified certain areas in the brain which relate to cognisance (or recognition of faces / voices), communication and command/control over our physical limbs. Similarly, in the way in which the Net is forming we have ways of cognisance (via avatars and images), communication (text, images, IMs, emails etc) and a command/control function in the coding which paths all those IF mouse is clicked, XYZ happens or WHEN text is input, insert into database type commands which appear in Boolean, Javascript, AS3, Squeak and every other OOP (object-oriented program) which makes up what we call the World Wide Web or Net.

So I started to make the connections between all the literature (including business psychology models) I’d read since childhood, my own observations of how the brain works in situ (including young children and spending time with my grandparents as much as daily interactions with people @ work and @ play), my father’s situation and my work on the Net and this realization sparked:

* WE HAVEN’T DISCOVERED CONSCIOUSNESS YET BECAUSE WE’RE LOOKING IN THE WRONG PLACES AND WITH THE WRONG TOOLS! What if it’s not via ECGs and MRIs alone?

Then the challenge becomes, “Well what model or framework can we build to detect it and guide us to finding the right tools?”

Through the interactions with my father it became apparent that he had cognisance of who I was and also of Elvis and Pavarotti when I plugged in his music headset. His facial expressions would change subtly but perceptibly. There was also moistness which formed in his eyes and showed up on his lashes. He could sense and was moved by the music, that’s how I interpreted this moistness. I can imagine how frustrating it must have been for him: an intelligent and articulate person who was in a vegetative state.

To the hospital staff, he was a patient number. To us, he was ORGANIC: our fellow journeyer through Life’s ups and downs, evolving and mutating along the way. He was the one who — together with our mother — taught us how to walk, talk, read, write, laugh, cry, imagine, be and a billion other shared experiences (good and bad). Just as he had comforted us, washed our faces and held our hands when we were relatively young and dependent, so it was our turn to hold his hand, wash his face and do whatever we could to comfort him. The sensation of touch was another way for me to gauge his state of being.

Here too I got the sense that he was conscious and aware he wasn’t alone and that we were with him. He couldn’t grip my hand back but occasionally there was a pulsation on the tips and it would become warmer.

After a few days, it struck me that whilst he’d lost functional ability of his communication, command/control, collaboration and coherence faculties I didn’t (and still don’t) personally believe that he’d lost consciousness entirely — only the consciousness as currently defined by medical information and the tools available.

What’s irrefutable is that we haven’t definitively found consciousness or its location yet. If we had, I’d probably have read about it in The Lancet, New Scientist, Wired, Nature, Scientific American, British Medical Journal, Neurosurgery Quarterly etc. (i.e., any of the specialist medical publications listed here: http://www.medic8.com/Journals/All.htm).

So I started to think about, “What are the core elements of consciousness then — if the medical one is incomplete? Maybe once we find the core elements we might be able to narrow down the zones within the cortex where consciousness is triangulated.”

Culture, I thought, must be in there somewhere. Each of us is born into a particular culture and that DNA inherited from our parents must contribute to our consciousness, its course and its shaping in our histories, here and now and futures.

My father had a sense of his own culture because when I spoke with him in Chinese, again there were those subtle changes in his face. When the nurse(s) came to follow through with their procedures and addressed him in English I could see that his face was expressionless. If I read a passage to him from a book on plants, that expression was different from when I read a passage from a historical Chinese novel. When he was listening to Pavarotti there was a glow to his face which was different from if I played a Chinese female singer from the 1940s and 1950s.

Anyway, more recently whilst tracking developments on the Net and the building of “The Global Brain” I realized that culture is a core component here too. We talk about Semantics and yet the definition of semantics means different things to different people (around the world, across genders, traversing cultures and educational / professional reference points).

To Tim Berners-Lee and the W3C it means a set of ontologies to help us classify data objects. To me, it means those ontologies PLUS taking into account cultural and perceptual factors like subtle nuances, double entendres, potential lost in translations, the differences between male-female communication etc.

I think also of the coherence component. We can have ontologies which stand up in their own right and yet are not coherent in the whole. So, for example, the logic of their classification doesn’t synch with another’s. Paris in RDF form is a location, a proper noun and a fictional character from Homer’s Odysseus but, presently, if we went on the streets and asked people, “What does Paris mean to you?” the answer would not be “Capital city of France / Paris Hilton / Paris, Prince of Troy.” John / Jane Doe on the street is more likely to say:

* It means romance.

* It means the Eiffel Tower / Sacre Coeur / the Louvre / La Rive Gauche (the Left Bank) etc.

* It means Sartre / Voltaire / de Beauvoir etc.

* It means an eye line that’s different / compact / elegantly distinguishable from London, New York and Toyko: fewer skyscrapers, more central planning.

* It means expensive / chic / beautiful / etc.

So is this set of classifications coherent with the noun set? No, it’s not. That’s because the adjectives set hasn’t yet been accounted for in the W3C design (I’ve accounted for it, though, in my model and algorithms).

Within the coherence component we also have to think about the clustering approach and whether the Bayesian tree filter approach is the optimal model for clustering. I would argue not (but that’s another post and some more emails between me and the MIT Collaboratorium team).

Once we crack the coherence component, the next ones to focus on would be consideration and creativity. What tools can be developed to harness or enhance those?

Again, if we compare the Net’s potential Global Brain with the actual human brain we can see that according to my model, if we can establish the definitive components and where they reside (cognisance, communication, command / control, collaboration, coherence, creativity, consideration and culture) we may pinpoint the holistic manifestation of consciousness itself.

That’s something good and positive to work together towards…….

I read this IBM news release and it’s definitely worth going over (and clicking through its various links) because at its crux it reveals how current Semantic Web pioneers’ attempts to build their version of a Global Brain with AI are likely to fail and won’t result in tools as smart or as proxy to the human brain as they’d like and think.

It’s notable that IBM differentiates between the need to simulate:

·      perception

·      multi-lateral processing

·      emotions

·      synaptronics

·      an alternative to the von Neumann bottleneck of purely processing words backwards and forwards

The lead of the IBM research project, Dharmendra Modha, notes:

“The vision for the anticipated DARPA SyNAPSE program is the enabling of electronic neuromorphic machine technology that is scalable to biological levels.  Programmable machines are limited not only by their computational capacity, but also by an architecture requiring (human-derived) algorithms to both describe and process information from their environment.  In contrast, biological neural systems (e.g., brains) autonomously process information in complex environments by automatically learning relevant and probabilistically stable features and associations.  Since real world systems are always many body problems with infinite combinatorial complexity, neuromorphic electronic machines would be preferable in a host of applications—but useful and practical implementations do not yet exist.”

For me, in certain ways, whilst the frameworks of the Semantic Web Stack are potentially enormously useful for sifting through all the billions of documents already online and interconnecting them with each other, that still doesn’t quite mean that the end-result will be a Global Brain whereby computers will be able to solve the world’s most complex issues: climate change, disease eradication, global democracy, universal education and poverty reduction.  

It will be a better-organized encyclopedia with annotations by multiple editors and library stewards is all. It focusses on ontologies and social graphs but not on multi-sensory discern of interpretation or synapses.

For a genuine Global Brain the IBM Synapse team’s aims are worth following.

Mohda also notes:

Synapses are junctions between neurons. In mouse and rat brains, there are roughly 10,000 times more synapses in the brain than neurons. Strength/efficacy/efficiency of synapses is subject to change (plasticity) as the animal interacts with the environment, and these synaptic junctions are hyothesized to encode our individual experience. The computation, communication, memory, power, and space requirements for representing brain in software or hardware seem to scale with the number of synapses. Thus, brain is much less a neural network, and more correctly, a synaptic network.

This is a very valid observation. I’ve worked previously with neural networks and AI models in asset allocation software. The results the system generated still had to be sanity-checked by humans — even though the algorithms were derived from them in the first place.

There is other contiguous and multi-lateral information that resides in human synapses and are not yet programmable into neural net, AI or SemWeb solutions.

IBM’s attempts to simulate these synapses is a “must follow”.

Perhaps the natural life cycle stage of the Web after the Semantic Web is the Synaptic Web, the Smart Solutions Web and the Seer Web BEFORE the Singularity?