Will neural networks revive and defunct areas of Philosophy?

Quantum NN are gonna unlock the mysteries of the Universe

Top kek

im trying to think of a joke with a robot and diogenis but i cant think of one. The pieces are all there... can anyone help me

Once they stop playing around with ANNs and start simulating real neural networks it will reveal how human intellect and compulsions of the mind function. Philosophy will be promptly buried after realizing it was all just mental entertainment.

Philosophy is the mozilla firefox of humanities, good luck reanimating a rotting corpse.

no.

They will be intentionally broken and disabled when they start spitting out politically incorrect redpills

>durr theyre like neurons so they'll be just like person

Neural networks are basically just a way to automate the feedback cycle for optimizing heuristics. They're a useful tool for areas of computer science for which "the correct answer" doesn't exist, but in terms of self-directing processes, it's about as impressive and philosophically intriguing as growing moss in a Petri dish.

>You have to already know the answer you’re looking for

Nobody knew what moves Alpha Zero was going to make.

obviously, but just moves aren't the goal, the goal is winning the game. so you train a model with the goal of winning the game and the network will provide moves which provide the best likelihood of reaching that state. read a bit into the theory first before getting excited about techniques which have existed for decades.

It's just automated statistics, or a mathematical model generated by automated statistics. If anything it'll be LESS enlightening since there will be less money put into actually figuring out what the equations should be.

Yeah well try and get Alpha Zero to play tic-tac-toe. It's not even something that it's possible to do at a conceptual level. Because it's not intelligent, it's just a fucking program that plays chess. Which is already over-personifying the physical processes going on.

NN are not enough to have a sentient AI.
NN are good to build intuition and logic but not enough to simulate empathy and things to like/dislike based purely on environmental/social interactions.
You won't be able to produce a philosophy bot with our current techs and we still don't know what we need to do to go there.

>Make philosopher AI
>Get scrapped for making analysis that hurt human feelings

Attached: 15658698545.png (772x499, 137K)

But everybody knew the answer to the question: score more points than your opponent

>glorified pattern matching and fuzzy model
>hurr this revlotiouns everting!

stop reading any news regarding the subject from general media with special retards award on topic going to Ars Technica for 'pretending we get it but really just sell you clickbait' excellence

Neural networks aren't actually very good replicas on how the human mind world. When we try to get them to act just like real neurons, they kind of sucks.

Big data is a big delusion. Stick to money ball. The feed-back loop between hiring and experiencing the benefit of the hire is simultaneously too distant and too abstract for machine learning to be in any way practical. It's not a game with rules.

>Big Data is a big delusion
That is true.

NNs biggest impacts will be in pornography and propaganda
prove me wrong.
(protip, you can't)

>A priest, a sentient AI and Diogenes walk into a bar.

Sorry, that's all I've got.

>goal is winning the game
I'm looking forward to see what bigger models can do when trained on broader tasks.

>Philosophy will be promptly buried after realizing it was all just mental entertainment.
A lot of 20th Century Philosophy already did this. I see a lot of people laboring under false assumptions discredited by Heidegger and Wittgenstein.
Good for someone working in natural language processing to know 20th century philosophy so as to not labor under false assumptions.

NN's biggest successes have been in image and audio processing/generation. So, yeah, pornography and propaganda sound about right.

in 10 years, pornography companies won't need to pay models. films will be auto generated. pornography may well even be custom generated per-person based on his or her personal kinks. it will learn what faces and figures you find most attractive.

in the propaganda sphere, it will be abused by intelligence services for psychological warfare

semi related, this is a fun read slatestarcodex.com/2018/10/30/sort-by-controversial/

You would not say this if you knew about the technologies you were talking about.

>But everybody knew the answer to the question: score more points than your opponent

No. It made moves nobody could comprehend making.

Go and chess are pretty closed, logic systems, so apt for machine learning.
Still, it feels like people will apply it to things that will be obvious in hindsight.

Not the original guy but I don't see why not. I'm not sure about the legality but with the huge amount of free porn online a startup could grab it and apply any sort of neural network based transfer so that the user could choose a face, voice, body type, etc...

>I'm looking forward to see what bigger models can do when trained on broader tasks.
neural networks aren't a solution. they excel at problems with clearly defined goals, like playing a game, analyzing images, and replicating documents. the vaguer the target, the less practical they are because you would need exponentially more data and time to train them.

will meme learning ever make meme science obsolete?
one of the greatest questions of our time.

>would need exponentially more data and time

I think that might be coming with more compute power and I think they will be able to "imagine" new data to train on based on the data available.

I don't have the attitude of this thread of assuming what they won't be able to do. I am aware they are hyped, slow to learn, hard to train, and narrow right now.

What if you had a 1080ti that was 100 million times more powerful than a normal one. What models could you train?