Live data from Hacker News

What is it like to be a bat? (1974) [pdf]

warwick.ac.uk

1–10 of 121 posts

Re: What is it like to be a bat? (1974) [pdf]

#4
> But bat sonar, though clearly a form of perception,is not similar in its operation to any sense that we possess, and there is no reason to suppose that it is subjectively like anything we can experience or imagine.

Philosophy aside, bat sonar is different from the senses we possess in an really interesting way. Our eyes have excellent spatial resolution (up/down, left/right), some rough depth resolution (from stereo), and no innate sense of speed. Our brain processes the signal to fake even better spatial resolution, infer more about depth (small vs far away) and more about speed (angle changes, among other things).

Bat sonar is completely different. Spatial resolution is poor. But they have first-class depth and speed information! They don't necessarily know where something is, but know exactly how far away it is, and how fast that distance is changing. One must suppose that their brains synthesize more spatial information from these senses, but that spatial information is still not going to feel reliable.

I'd love to be able to experience that for an hour. To live in this world where distance and speed are primary senses, and cross-range information is much fuzzier. What an incredibly different way to see the world.

Re: What is it like to be a bat? (1974) [pdf]

#6
I think the closest we might get to actually learning what it is like to be a bat is by unsupervised learning + a bit of manual labelling on top. A neural net could learn their states and dynamics, by training on a million hours of bat recordings. These representations will already encode bat states and values, so we just need the bridge to human language, which is easy to build with a pretrained language model.

This approach works for any species, neural nets can do it because they can do unsupervised learning. I bet we'll see pet translator apps popping up. Maybe we can monitor the environment by listening in on animal chatter.

Re: What is it like to be a bat? (1974) [pdf]

#7
One excellent book on consciousness is Mcginn's The Mysterious Flame [1]. I knew consciousness is a hard problem, but the book made it clear how baffling and utterly mysterious it is. I am still absolutely flabbergasted when I think about it (how on earth does consciousness arise from material "meat"?! [2] Where are pain and color and subjective experience really located in the material universe?). It also made me skeptical of people who think AI will be sentient [3] while we are in the complete dark about consciousness in biological organisms.

[1] https://www.amazon.com/Mysterious-Flame-Conscious-Minds-Mate...

[2] https://www.mit.edu/people/dpolicar/writing/prose/text/think...

[3] https://twitter.com/lexfridman/status/1653051310034305025

Re: What is it like to be a bat? (1974) [pdf]

#8
I've read this before. That's why I think the question "do animals have consciousness?" is meaningless, because "consciousness" usually implies "like ours."

They have something that probably bears some relationship to ours. Some birds have a "theory of mind" where they know what you know, e.g. whether you saw them hide the food.

It would be possible (maybe someone's already done it?) to enumerate the N tests of "consciousness," where if an organism has all those N, then it's "conscious." Someone would object "oh, but humans can do so much more than those!" and that's true. So if you increase N enough, only humans are "conscious."

Re: What is it like to be a bat? (1974) [pdf]

#9
post #6

I think the closest we might get to actually learning what it is like to be a bat is by unsupervised learning + a bit of manual labelling on top. A neural net could learn their states and dynamics, by training on a million hours of bat recordings. These representations will already encode bat states and values, so we just need the bridge to human language, which is easy to build with a pretrained language model. This…

The point is that no amount of external modeling can give you any knowledge of “what it’s like” subjectively.

Re: What is it like to be a bat? (1974) [pdf]

#10
post #7

One excellent book on consciousness is Mcginn's The Mysterious Flame [1]. I knew consciousness is a hard problem, but the book made it clear how baffling and utterly mysterious it is. I am still absolutely flabbergasted when I think about it (how on earth does consciousness arise from material "meat"?! [2] Where are pain and color and subjective experience really located in the material universe?). It also made me sk…

> how on earth does consciousness arise from material "meat"

Let me try. It is a way to perceive with goals in mind. Nothing special, just perception, future reward prediction conditioned on current state, and learning from outcomes.

The whole specialness of consciousness is that it carries inside not just our external world, but also our plans and goals. So in this place where they meet, and where there are consequences to be had for each decision, this is where consciousness is. [*]

I support the 5E theory of cognition - embodied, embedded, extended, enacted, and enactive cognition. I think you need to look not at the brain but the whole game to find consciousness.

[*] Consequences for an AI agent could be changing its neural network, so it updates its behaviour, and changes in the external situation - the agent might not be able to backtrack past a decision they take.

Post reply on HN