/sci/ - Science & Math » Thread #11438849

22KiB, 638x330, IMG_3174.jpg

View Same Google iqdb SauceNAO

Anonymous Wed 04 Mar 07:02:31 2020 No.11438849 View Reply Original Report

Quoted By: >>11438897 >>11438906 >>11438908

What is the difference between a machine learning neural network and a physical brain? Would you be able to implement the same logic of machine learning to say an insect? Or a rodent?

If the technology or understanding for neural networks or A.I. were to advance, would it give us a clearer idea on how the brain or consciousness works?

t. brainlet

Anonymous

Anonymous Wed 04 Mar 2020 07:38:32 No.11438897 Report

Quoted By: >>11438906 >>11438908

>>11438849
A brainlet response would be - not a lot. That's because ML neural nets are based on the physical brain. They activate only when needed and can also self correct.
The harder answer would be - we don't know. Primarily because neurologist still don't know how neurons exactly function. Papers have been written about the general overall idea about their function and but don't know precisely how billions of neurons come together in parallel to form a living brain. It's highly complex. Hell, even a mouse brain has billions of neurons and no one has come close to mapping each and every neuron. Also with how complex the whole brain is, some scientist speculate that neurons maybe little computers themselves not dumb activators.
>If the technology or understanding for neural networks or A.I. were to advance, would it give us a clearer idea on how the brain or consciousness works?
No. Maybe not for a long time. Our primary hurdle is currently ANNs and the brain itself. It's very expensive to simulate billions of neurons on a computer. Also it's extremely difficult. People have been looking into alternative physical ways of simulating it. Also we don't understand how the human brain works such as sight, memory let alone something as complex as consciousness.

Anonymous

Anonymous Wed 04 Mar 2020 07:47:47 No.11438906 Report

Quoted By:

>>11438849
>>11438897(2/2)
To add more, consider for the fact that the brain has a power equivalence to that of a light bulb while computers pull megawatts of power but still lag behind in terms of general intelligence.
To give some perspective, here are some fundamental problems with neural nets till suffer from
1) They require too much data to give reasonable accuracy. Consider that the human brain, once seeing an apple, will always currently classify the apple in almost all cases. While a similar system cannot identify a tilted bus.
2) Also state of the art AI systems are still highly specialized. A system extremely good in vision performs badly at simple tasks like sentiment analysis from task or games such as chess

Anonymous

Anonymous Wed 04 Mar 2020 07:49:08 No.11438908 Report

Quoted By:

>>11438849 (OP)
>>11438897 (You)(2/2)
To add more, consider for the fact that the brain has a power equivalence to that of a light bulb while computers pull megawatts of power but still lag behind in terms of general intelligence.
To give some perspective, here are some fundamental problems that neural nets till suffer from
1) They require too much data to give reasonable accuracy. Consider that the human brain, once seeing an apple, will always correctly classify the apple in almost all cases. While a similar system cannot identify a tilted bus.
2) Also state of the art AI systems are still highly specialized. A system extremely good in vision performs badly at simple tasks like sentiment analysis from text or games such as chess

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches