"There is something special about humanity, and we should celebrate that."

On LinkedIn I ended up posting a lot of comments on the subject of the (unlikely soon IMO) imminent prospect of AGI at Human levels and beyond being achieved very soon, countering disparaging comments from Peter Norvig et al that…

“Insofar as sceptics remain unmoved by metrics, they may be unwilling to accept any empirical evidence of AGI. Such reluctance can be driven by a desire to maintain something special about the human spirit, just as humanity has been reluctant to accept that the Earth is not the centre of the universe and that Homo sapiens are not the pinnacle of a “great chain of being.” There is indeed something special about humanity, and we should celebrate that, but we should not conflate it with general intelligence"

See… https://www.noemamag.com/artificial-general-intelligence.../

Here are a selection of my argument the final sentence… “…there is something special about humanity, and we should celebrate that, but we should not conflate it with general intelligence“. Since that something “special” represents really difficult problems in the development of AGI!

*****

1/ Whilst LLMs are so complex it’s difficult to peer inside the black box. Yet we engineer the processes of Transformer tech, and understand how they work, we understand the assignments of activation functions in Deep Learning. These systems in the Token Processing functions simply do not evoke reflective cybernetic feedback loops necessary in Metacognition. To claim System 2 reasoning is present under emergence, based on behaviour is not scientific. You need to falsify the alternative hypothesis that it’s mimicking. The only way is to look inside the black box and find those pesky Cybernetic Metacognition loops!

My bet is they don’t exist and these systems are very sophisticated stochastic parrots, produced at huge environmental costs.

2/ We know Neural Plasticity is a fact. It is hypothesized to occur during Eureka moments (sometimes after a period of relaxation after deep reflection). The creation of a new cognitive substrate enables such cognition to grow in ways that can enable human cognition to have Oracle Machine facility. I trust you seen/read my link on Turing/Oracle Machines. G?del is on the record as saying…

“Either mathematics is too big for the human mind, or the human mind is more than a machine”.

This quote suggests that the human mind is capable of more than just mechanical processing and that it can comprehend concepts beyond the scope of machines.

BTW I posted the other day on a paper by Solomon Feferman whereby he argues the uniqueness of humans having Free Will by G?dellian arguments. I’ll post a link to that for you.

See… https://www.dhirubhai.net/.../martin-ciupa-76418b17_title...

3/ Embodiment (rich in sensory access) is a very good and perhaps necessary step to enabling AI to develop its cognitive causal reality model and common sense. Try to learn about the world from language alone is not enough.

But to really understand the world in the ways humans do may need evolution over billions of years with trillions of simultaneous ancestors struggling to survive and thrive. By doing so we are likely to become progressively fit to nature. We may have gained our Causal Reality Model and Common Sense intuitive awareness this way.

Some humans have remarkable intuitive sense. I’ve written several related posts on the subject of #Mathematicalintuition. You might want to look those up.

I put one link here as a starter…. https://www.bbvaopenmind.com/.../ramanujan-the-man-who.../

4/ I’ve served as a CAIO of a unicorn Neurotech company and I know a thing or two about concepts of brain simulation and uploading.

It’s complete and utter nonsense of a fanciful sci-fi hope.

To make a contemporaneous image of not just the connectome, but down to each of the activation functions of each of the 10k synapses per neuron of the 86 billion neurons in the average brain would fry the brain by fMRI (that doesn’t record the activation potentials any way).

You’d have to model also the EEG field that is thought to help coordinate consciousness as well as the Neurochemistry flux.

Try to do this in snapshots over time wouldn’t work, you end up with a jigsaw puzzle of a moving picture. (as I mentioned the brain changes).

Give up on that silliness, please

5/ AGI may be possible, probably with hypercomputing and neuromorphic technology.

There are a lot of AI folk filled with an ideological ambition of TESCREALism and are at risk of falling into a Moloch trap.

I’ve posted on that also. You can look those up or Google them!

.

What we should be doing is building #AI4good. Aligned to UN sustainability goals. Not pursuing an ideological goal that has alignment and containment concerns.

We can do that. We should do that.