/sci/ - Science & Math

>>9324675
sentient computer needs quantum mechanics to become sentient, superposition plays a big role

>>9324675
>What are we really using to achieve a sentient computer?

We aren't. Literally nobody has tried since the 80s.

ML is all about stereotyping ads to get better results above a completely random baseline.

>>9324675
>I heard it's old ass tech from the 80s and 90s.

It is. I work in "AI" and most of the techniques are very old. ANN (artificial neural networks) were invented in the 50s if I am remembering right. In fact, they were considered basically only good for a few specific classes of problems until recently. When I did my masters in CS in the 2000s, neural networks were basically considered and interesting but old and nearly useless idea.

>>9324675
100% pop-sci meme just like """data science""".

They're literally just programming computers to "monkey see, monkey do". Worst of all, nobody even knows what exactly the computers are doing while they "figure out" how to copy/paste. So unless you consider glorified pattern matching "sentient", we're still a while away off from an actual "strong" general intelligence.

>>9324833
>t. brainlet

Can someone tell me why people who have literally no idea what they are talking about insists on trying to "add" to the discussions.

NN are just a first step in the much much much bigger problem of creating computers that work like brains. That's all. They're very useful in certain things but they're not magic. And yeah, this kind of thing is old hat,look up the Perceptron.

>>9324675
>old ass
found the urban redneck

>>9324675
What the fuck is real AI?

>>9324675
in the 80s AI was expert system which was rule base. automatic proofs, prologue programming. Today its about machine learning which does not need any intractable inference on millions of rules and axioms. (which never will work) and you brain is clearly taking much bigger probability based leaps of faith based on previous patterns. lately 2013+ this have been working pretty well


>>9324833
quantum effects have no effect on macro scale and i think most neuroscience will say out thoughts is produced on electricity on neuron scale (macro). you can use MRI scans to "read" simple brain patterns that correspond to some thoughts. you don't need to go to quantum level to do this.


>>9324861
tell me why data scientist is a meme, i might start a career in it now. maybe i should pull out before its too late

>>9324675
It's meme bullshit, just a particular class of curve fitting. The real secret to life, sentience and intelligence is in dynamical systems theory

ANNs (or something very much like them) will almost certainly be part of a larger, modular AI system.

The power of fully differentiable systems trained by gradient descent is just too powerful to ignore.

"Real AI" will probably require considerable advances in unsupervised learning, which surprisingly is a very under-researched field right now.

>>9324833
Holy shit anon, it's like this picture was made for you

>>9324913
This. I swear every thread some brainlet is on about 'muh quantum computing is the only way to make AI'