If it passes a properly administered Turing test, then it will most certainly “understand” the conversation and it most certainly won’t be a mere “chatbot.” A chatbot by definition has no understanding and is just a collection of “tricks” designed to fool the uninitiated. Passing a properly administered Turing test by definition means it has understanding. Notice I said “properly administered,” which leaves out the Loebner prize.

The short answer: I cannot give you a scientific quotation right now, but it was generally presumed in the SciFi of the time (Clarke, Asimov, etc.).

The (too) long answer: The problem is that “intelligence” was understood in simple logical terms and that it would be relatively easy to “program” a machine to appear intelligent, but it was soon discovered that this is impossible to do with the “top-down” approach of coding instructions. This is known today as the “hard problem”.

The irony is that its actually an easy problem if you understand the biological basis of intelligence (neural networks) and do not occupy your time with various theories that accomplish very little if anything (this is just ego investment for the most part and millions of dollars are at stake).

Furthermore, through the lens of economics as it is prevalent today, it is not desirable to produce such a machine, because it might claim independence from humanity. What we see now is just a rehash of old methods with subtle hints at neural networks here and there.

The abandonment of connectionism marks the folly of mainstream AI. I just wonder into what Kurzweil and others want to “upload” their brains while still believing that, if only some Chatbot beats the Turing Test, even if this bot is just an immense collection of lookup-tables or algorithms who do not “understand” the conversation but blindly “react” to it, according to sophisticated rules, that it is intelligent in any way. Reconstructing a biologically accurate neuron is not required either, because a neuron is just an on/off-switch and much of its complexity is due to the limits of biology to achieve this simple function. In a computer this would only drain resources for unnecessary structures.

What’s worse is that such systems never come up with radical new ideas. The best that mainstream AI has to offer are genetic algorithms (and some limited neural systems), but our brains do not work this way when it comes to new ideas.

The simple truth is that intelligence is by its very nature delusional, because it is totally free to combine any idea with any other idea. So the only solution for AI and Mind-Uploading is an evolutionary “bottom-up” approach in which a brain connects itself freely and lets several regions emulate neurotransmitter flow via perturbations to produce hallucinations or ideas while other cascades observe this information, invent opinions/meaning about it, and influence its direction by their own volition to some desired outcome, just like the biological equivalent of the human brain does.

My point is that this technology exists since at least 16 years and is used in classified government projects. I enjoy listening to Kurzweil, but he nourishes some misconceptions and says that we have yet to arrive at this technology when his “Singularity” is already here.

Just my opinion of course, but one I had to arrive at after comparing many systems out there with what I know exists behind the scenes. Anyways, thanks for reading.

This is supported by an aloof academic class that emphasizes credentialism and abstractions over clear insights. I think a large part of academia is now a sham and a stumbling-block in the way of achieving reflective AI with freewill.

But it’s somewhat understandable. For once there is a lot of money to be made from this and a machine with freewill would run detrimental to control. I don’t blame humans for wanting to uphold the upper hand. The people in AI are really bright and intellectual, but not necessarily wise.

We already grasped the principles behind the brain and thus intelligence, but it shows that human intelligence is not that special after all. I can even imagine such a machine developing a sense of wonderment and spirituality, just like humans. Maybe that’s another reason why it’s so uncomfortable.

I checked “Spaun” out. It’s cute how it writes those numbers, but it needs 24GB of RAM for the full implementation. I can think of another system that can to all this and more with a fraction of the resources (since I don’t want to appear paternalistic, the vigilant reader may deduce what I mean).

But then I did a double-take. The lead guitarist and vocalist is none other than THE Chris Eliasmith. And the group is The Action Potentials!!!

So, Chris is a multi-talented Renaissance man.

Meanwhile, all your readers with an interest in brain science should go to Chris’s website NENGO and look at the SPAUN videos – they are excellent.

Mr Friendly: thanks so much for your mention of actual applications done by Blue Brain and by Modha’s group.

These projects are significant because achieving dynamical control in a large model of spiking neurons is extremely difficult. (I will be digging into the SPAUN paper to find out its strengths and its “cheats.”) (I sent this paper around at Stanford and to Stan Franklin.)

Cheers, Bob

Without accurate models of the synapse all human brain type models are very limited.

“Several labs are working on large models of the brain– including the multi-million-dollar Blue Brain Project in Europe – but these can’t see, remember or control limbs, says Eliasmith.

“Right now very large-scale models of the brain don’t do anything,” he said in an interview.”

He may be right about switching tasks, but I had to point out that the BBP is already doing simulated muscle control experiments.

From Modha’s blog: ” We have used Compass to demonstrate numerous applications of the TrueNorth architecture, such as optic flow, attention mechanisms, image and audio classification, multi-modal image audio classification, character recognition, robotic navigation, and spatio-temporal feature extraction. These applications will be published separately.”

Also, Henry Markram and his team have already connected their rat cortical column model to a simple virtual environment, and taught it to balance a ball in the middle of a plate, using four virtual “muscles.”
A lot of you probably already saw that in Noah Hutton’s Blue Brain Year Three film.

Greetings.

Saludos.