[normalhuman]

The idea of "embodied AI" has been around for some decades. It is reasonable that, from a practical engineering perspective, creating "human-like" intelligence becomes more feasible if this intelligence is embodied in the same physical possibilities and constraints as a typical human.

What I find somewhat ironic is this: the author mentions working with Stephen Hawking, an amazing man who produced incredible intellectual work and enriched our understanding of reality while being almost incapable of any physicality.

If we apply current scientific theories (physics, chemistry, biology, etc) to this cell network and physical machinery, we quickly find our way back to symbolic manipulation. What are cells if not computational nodes that exchange messages?

A much more reasonable hypothesis for what is missing is contained in the text:

"A human cell is a remarkable piece of networked machinery that has about the same number of components as a modern jumbo jet[...]"

Maybe we just haven't reached the level of complexity needed for human-level AI. A hint that this might be the case is that the current excitement with ML seems to be fueled by algorithms that were mostly known by the 80s (sure, with lots of recent incremental improvements, but no new big idea). What made a difference was the computational power and datasets that became available in the 2010s. I suspect the next leap will be of a similar nature. "More is different".

[dustfinger]

> Maybe we just haven't reached the level of complexity needed for human-level AI. A hint that this might be the case is that the current excitement with ML seems to be fueled by algorithms that were mostly known by the 80s (sure, with lots of recent incremental improvements, but no new big idea). What made a difference was the computational power and datasets that became available in the 2010s. I suspect the next leap will be of a similar nature. "More is different".

Regarding the nature of complexity and the notion that "More is different", I am reminded of the emergent behavior of vivisystems [1] as described in Kevin Kelley's book Out of Control [2] -- an insightful exploration of the emergent behavior expressed by complex self sustaining systems. If you have not read Out of Control then you might want to put it in your reading queue. I found it highly engaging and thought provoking.

[1] https://www.everything2.com/title/Vivisystem

[2] https://kk.org/outofcontrol/

[visarga]

I like the alternative phrase 'Quantity has a quality of its own'

[antisemiotic]

>What I find somewhat ironic is this: the author mentions working with Stephen Hawking, an amazing man who produced incredible intellectual work and enriched our understanding of reality while being almost incapable of any physicality.

Not from birth though. I wouldn't dismiss the impact of being able to interact with the world during childhood so easily.

[mannykannot]

Helen Keller is a better example. It is also the case, however, that her brain, like anyone's, had been shaped, through evolution, by eons of interaction with the environment.

[jcims]

To me embodiment is important because it acts as the root in the hierarchy of reference frames from which you can model the real world. So whether or not the body has a particular set of physical capabilities isn't as important as the relationship it has with the rest of the world around it.

[wpietri]

This strikes me as appealing but dangerous:

> What are cells if not computational nodes that exchange messages?

We evolved in a context where we survived by exchanging messages with other "computational nodes". So it's very tempting to see everything through that lens. But I think it's a mistake to see cells as "really" just like us. As Box said, "All models are wrong, some models are useful." We shouldn't forget that cells are really cells, and we see them as analogous to familiar things because the world is too big to represent directly in three pounds of meat.

[gmueckl]

Are you sure that the parent was comparong humans to cells? It strikes me more as this: "human intelligence is run by a network of nerve cells which are complex but could be modeled in all of their conplexity". So if you could model enough of these, you would get a functional replica of a human brain.

[ptidhomme]

> What I find somewhat ironic is this: the author mentions working with Stephen Hawking, an amazing man who produced incredible intellectual work and enriched our understanding of reality while being almost incapable of any physicality.

From my point of view, what matters is not physical interaction but rather physical perception. May it be visual, auditive, tactile etc. it all contributes to build the "model of the world" you carry in your whole body.

I suppose Hawking could still largely perceive his environment.

[sumitviii]

>while being almost incapable of any physicality

Only for the later (even if major) part of his life. One could argue that the physicality he was capable of in the earlier stage of his life helped him gain a solid understanding of physics on our level.

But there is a better argument to be made: Modern Physics is an entirely different beast. To understand the underlying reality you have to be, in one sense of the word, be detached from the reality. And being paralyzed can be said to be one of the many things that let him have experiences unlike that of any almost any other human being. His earlier life let him have a solid footing on the dynamics of our level, and his later life allowed him to depart from it, in a direction in which he was propelled by his intellect.

I would argue that if Stephen Hawking was a sports instructor (or even a programmer for a sport software), his lack of physicality would have worked against him. But if you showed me someone with a severe physical disability who writes great code for sport software, I would revert to my first argument.

(I just explain what's in front of me! :P)

[JoeAltmaier]

I think the jumbo-jet argument is off a little - maybe 5 or 6 orders of magnitude? Anyway, a vivid analogy.

[jackhack]

If you mean the jet airplane is the simpler of the two, I would agree. The famous Roche Biochemical Pathways chart of just those mecahnisms that are understood illustrates this perfectly: http://www.expasy.ch/cgi-bin/show_thumbnails.pl

[JoeAltmaier]

Right. Proteins alone, there are something link 10^9 in one cell. Plus all the 'parts' the proteins deal with (many more)