Douglas Hofstadter: Artificial neural networks today are not conscious

https://www.economist.com/by-invitation/2022/06/09/artificial-neural-networks-today-are-not-conscious-according-to-douglas-hofstadter

by tosh • 4 years ago

53 16 4 years ago

economist.com

daniel-cussen2• 4 years ago

That's true, they're not concious.

But they're not properly contained either.

pyinstallwoes• 4 years ago

What does this mean?

daniel-cussen2• 4 years ago

An AI is not that different from a virus. You have to stop it from getting out, put it into a neat little black box that is not only impenetrable from the outside, but impenetrable from the inside, with a tiny little peephole for you to view it.

stevesimmons• 4 years ago

Duplicate of the thread at https://news.ycombinator.com/item?id=31689840

witheld• 4 years ago

One of them is it gonna think about itself, someday. But probably not on a non-analog computer.

Communitivity• 4 years ago

The first computer to think about itself will also probably not be a 100% non-organic computer, but either a 100% Wetware computer, or more likely a hybrid. A wetware computer is a computer made with organic parts. There are efforts underway to create organic computer components, and they've made slow but non-zero progress.

Two companies working in this area are Cortical Labs [1], and Koniku [2]. I am associated with neither.

Two past efforts I've seen are a design for a computer using a fungal network [3], and organic neural networks in fabrics [4]. A lot of work on chaotic computer, such as organic computers, was pioneered by William Ditto, at Georgia Institute of Technology [5].

[1] https://corticallabs.com/

[2] https://koniku.com/

[3] https://royalsocietypublishing.org/doi/10.1098/rsfs.2018.002...

[4] https://www.science.org/doi/10.1126/sciadv.aba1178

[5] https://labs.sciences.ncsu.edu/appliedchaoslab/william-ditto...

jdonaldson• 4 years ago

You have to be crazy to think awareness requires wetware. Silicon is going to zip past every possible measure of it in the next 5 years.

shreyshnaccount2• 4 years ago

I don't think anyone who knows what's going on has claimed otherwise... Or if they have idk about that

kgwgk• 4 years ago

The hype side: https://www.economist.com/by-invitation/2022/06/09/artificia...

Mo3• 4 years ago

From the same damn site too. sigh

mirekrusin• 4 years ago

The easy way to answer this question is to plug tiny extension to human brain so it grows connections. If it takes over parts of the brain, just iterate on step 1. You can cut off wetware parts as step 2 I guess. Are you going to be left with normal person that transferred to cyborg?

grumpopotamus• 4 years ago

It's unclear if we can ever know. How can you tell if the neural net is having a subjective conscious experience, or if it's just outputting words in the dark? Clearly they are already capable of very sophisticated communication, so if they're not conscious now, how will we ever know if the light bulb turns on? I can't see the part of tfa that's behind the paywall so not sure if this is discussed.

YossarianFrPrez• 4 years ago

We might only know for sure if and when we figure out the mechanisms that generate consciousness in brains.

In the mean time, it seems unlikely to me that matrix multiplication conducted via electro-atomic / quantum (and not molecular) effects will be the foundation of silicon-chip based consciousness.

bawolff• 4 years ago

Clearly they are already capable of very sophisticated communication

I don't think that's true.

Anyways, how can you tell if the pen on your desk is concious? If we cant answer that simple question, we shouldnt worry about the complex systems.

dark_star• 4 years ago

https://web.archive.org/web/20220610131709/https://www.econo...

Douglas Hofstadter: Artificial neural networks today are not conscious

Related Stories