What is intelligent life? Portia Spiders and GPT | Hacker News Reader

What is intelligent life? Portia Spiders and GPT

aeon.co

Ever since reading Children of Time, have never thought of Portia Spiders the same way again. Read it back to back with Blindsight, and it really shifted my view of consciousness.

And even current societies concepts of male/female. In Children of Time the Portia Spiders are Female dominant, and eat the males. Scale that up to an intelligent society and it was pretty interesting perspective change to see female characters debating if they should go hunt some males for dinner. And later discussing 'rights' and "of course males can't be equal, can't they just be happy if we let them live.."

I hate spiders (yes, I know the value they provide, I can't get over how much they creep me out) but I loved the Children of Time series and also I just re-read A Deepness in the Sky (Vernor Vinge) which is also a favorite of mine. I'm not sure I could handle a movie of either but the books are great since I can forget they are "spiders".

Well I loved A Deepness in the Sky and Blindsight so now I'm adding Children of Time to my ever growing list.

A very nice book indeed, and i also recommend the following tomes

I suspect that A Deepness in the Sky would work best as animation, to emphasise the same plot point that had the spiders using familiar human nouns and only reveal the graphical truth at the same time in the story we find out the literary truth.

(I vaguely remember something like this happening with a monster's POV section in Schlock Mercenary, but that's a long web comic and I can't remember when in it's history to look for).

The arc starts here: https://www.schlockmercenary.com/2001-10-01

It switches POV here: https://www.schlockmercenary.com/2001-10-22

Yup, that's the one I had in mind.

interesting connection between the mindrot from that book and the warfighter alertness link posted today that uses IR light to target drug activation in specific parts of the brain:

https://news.ycombinator.com/item?id=40707093

https://en.wikipedia.org/wiki/A_Deepness_in_the_Sky

And here is the obligatory Children of time mention. Beer shot

This is a wonderful essay.

I've always found the term AGI confusing. For example, how general does it need to be to qualify and what specific cognitive capabilities does it need to exhibit? My gut feeling has always been that it's not a helpful guiding star for AI research.

It seems better to be led by the problems we want to solve as that makes it easier to define success and generality is still beneficial since you can more efficiently solve more problems with general solutions. What this essay solidified for me is that what we call intelligence is really just a set of tools that allow us (humans) to solve a certain collection of problems. We mistakenly believe all those tools are cognitive, but really some are just evolved responses and instincts.

If you enjoy this topic I also highly recommend "A Brief History of Intelligence" which goes into quite a bit of detail, is very readable, and ties in directly to the near term future of what intelligence will mean in our world. Really a very good book!

Reminds me of BEAM Robotics from late last century. I was sort of amazed by the little kit I put together an watched it follow the sun. GPT is sort of amazing also but I am getting tired of listening to people (machines) talk.

https://en.wikipedia.org/wiki/BEAM_robotics

"Instead of a measurable, quantifiable thing that exists independently out in the world, we suggest that intelligence is a label." Thought experiments lead me to the conclusion that the same is true of consciousness. I make the analogy to how the essay, What Color Are Your Bits? which argues that copyright is "out of band" (my summary) from the actual bits of data. I think the property of a system having consciousness is, like copyright of data, neither intrinsic nor an epiphenomenon but rather at least partially a status of how we regard it.

The thought experiment goes thus:

Assume it's possible to simulate a conscious brain with a deterministic program. (Already a big ask for some, but this is one of the axioms of my argument so bear with me.) Assume that if embodied the simulation would be indistinguishable from the original person, but also if not embodied it could be interacted with in a simulated environment.

If it is deterministic, that means if you re-ran the simulation with same data , inputs, and timing, all would proceed identically. Is the simulation conscious again while replaying it along a fixed path?

Suppose you memoized portions of the computation. How much of the brain could you memoize before you no longer consider it conscious?

Suppose you could, instead of memoizing "in breadth", instead use memoization to skip steps of the consciousness without changing the outcome. How many in-between states could you gloss over while still considering the simulation conscious?

Suppose you divided the simulation among multiple computers, without changing the outcome. If you accept the original premise, you may have little trouble accepting a distributed system with a fast network is also capable of hosting a conscious mind. However as we said this replay is deterministic, there's nothing stopping nodes from substituting internal playback of pre-recorded network packets for actually communicating on the network.

Are you prepared to call it a simulation of consciousness if a "distributed system" of nodes each internally simulating all of it, in portions, while all remaining silent and not sending any network packets between them?

My point with all of these variations is not to say which is or isn't conscious, it's to argue there is no clear dividing line. One could come up with moral arguments about which of these scenarios it's unethical to keep trapped in this experimental setup versus which are the mere image of a thing and not the thing itself, but that's my point: it's only a moral/ethical dilemma and not a physical or informational state change between non-conscious and conscious. The "universe doesn't care;" consciousness is not a conserved property.

"Assume it's possible to simulate a conscious brain with a deterministic program."

I think this is touching on the 'philosophical zombie' arguments.

If something is crated like a robot/ai, that is completely indistinguishable from a human, is that thing even possible without having some inner subjective experience? It seems like people fall into two camps, if it is a machine, then of course it is a zombie, no inner life. The second camp that says the machine is conscious. The problem is from outside, you can't really prove it either way because the premise is that it is indistinguishable.

So, I think in your example, it is interesting extrapolation, the 'invention' is indistinguishable and deterministic, now what if we split it up, compress it, etc.. when would that inner experience go? or would it end at some point?

Of course, if a consciousness could be played back and forth like this, then that would be argument for GPT to have some consciousness .

> I think this is touching on the 'philosophical zombie' arguments.

As well as Buridan's ass: https://en.wikipedia.org/wiki/Buridan%27s_ass

These ideas are explored in Permutation City, by Greg Egan.

https://en.m.wikipedia.org/wiki/Permutation_City

Some of those thought experiments were interesting, but I still landed on "That's still a conscious system" for all of them. I expected you to go much further with the permutations.

Reply on news.ycombinator.com