-3
$\begingroup$

Brain waves work at most 50 operations per second, and the fastest synapse fire rate is not very different from this speed(up to 100 times per second). However, today computers operate in billions of operations per second(ghz), and computers working at terahertz speeds are in development. Besides that, brain and nervous cells are made of organic matter and organic matter is made of non metal elements. And you might know that non metals aren't good conductors of data and energy.

So AI systems could work billions, trillions of times faster than the human brain. Does that mean that they would think and process information much faster than a human can? Would their cognitive abilities be much faster as well? Just because they are faster mean they are smarter too?

$\endgroup$
  • 3
    $\begingroup$ This question seems rather muddled and confused. It assumes that there is a common measure of 'smart' that applies to both computers and brains, then spirals wildly, finally circling back around to wonder if perhaps such a measure might not exist after all. Consider refining this question in the Sandbox while doing a dollop more research on possible metrics for 'smart' $\endgroup$ – user535733 Jul 9 '18 at 14:35
  • 2
    $\begingroup$ strictly speaking also chips are made of non metals... $\endgroup$ – L.Dutch - Reinstate Monica Jul 9 '18 at 14:58
  • 2
    $\begingroup$ Hardware isn't everything; the quality of software also matters $\endgroup$ – nzaman Jul 9 '18 at 15:00
  • $\begingroup$ You seem to be asking several different questions some of which would contradict the others, I think user535733 was right to refer you to the Sandbox for a working over before we have another crack at it. $\endgroup$ – Ash Jul 9 '18 at 16:21
  • $\begingroup$ This question, at least to some degree, is a duplicate of Our brain is a very powerful signal processor, what's stopping us from creating computers to emulate (disclaimer, the link directs you to my answer). $\endgroup$ – JBH Jul 9 '18 at 17:15
4
$\begingroup$

I fear that the answer to your question is - we do not know. We do not really know how the brain does stuff in detail. Yeah, we know "Okay, seeing happens in that area, memories are mostly located there, and if we cut that out, he is a ditz!", but exactly how the brain manages to do things is up to alot of research.
So how should we know if an AI with a similar number of computing nodes is "faster"? (Ignoring the fact that "faster" is maybe the wrong word here, the brain has to handle our complete "OS", if we want to call it that)

Besides that, brain and nervous cells are made of organic matter and organic matter is made of non metal elements. And you might know that non metals aren't good conductors of data and energy

It seems that you mistake nerves as electrical conductors, but they are electro-chemical conductors. They do not convey data or energy, just signals.

There is another point to it: The brain is built to deal with its tasks, namely keeping us alive. It is a specialized tool that does its work (mostly) extrmely well. If you build a digital brain, it has to learn what to do, or it would be just a fancy calculator.

$\endgroup$
0
$\begingroup$

The answer is "maybe," but its not really measured in neurons at all.

There's 100 billion neurons in the human brain, give or take. A modern AMD Epyc 32 core processor weighs in at about 20 billion transistors. So 5 of them is on par with the brain, but we're clearly not at Artificial General Intelligences (AGI) yet!

So a transistor isn't a neuron. What can you do? Well, you can look at IBM's TrueNorth chips, which are designed to emulate the synapses of a brain. At the moment each chip has 256 million synapses. Estimates for number of synapses in the brain vary widely, but 150 trillion isn't a bad number. That means a mere 600 TrueNorth chips has the same number of synapses in the brain.

But what are you going to do with it? I can hand you 600,000 TrueNorth chips today, this very moment. Well, actually I can't, but IBM can. We can hook them together and have your neuron/synapse count today. They are very low power, so this would be easier than constructing many larger and more power hungry supercomputers. But does that help? When it comes to AI, it's the organization of the neurons that really matters. How do you configure the synapses. To this day, it is still unclear whether a computer can ever truly be more intelligent than an organic brain. The transhumanist in me wants to say it can, but science is still pulling back the veil on what it means to think. We can still be surprised.

By analogy, an empty hard drive, capable of storing 16,000,000,000,000 bytes of data is still just an empty platter. It's what you put in it that decides whether your hard drive is a collection of wisdom, or the world's largest collection of cat videos (Disclaimer: Cat videos may be the greatest source of wisdom on the planet. Disclaimer/Disclaimer: My cat may have been staring at me with murder in his eyes until I added that disclaimer)

If you really do want a numeric path towards AGI, I recommend reading up on Sentience Quotient (SQ). It was an approach which looked at the computational capacity of a brain versus its mass, and drew some conclusions about what might be required to think like a human. However, it's just numbers. The real question is can we make the patterns.

$\endgroup$
0
$\begingroup$

Provided a good definition of "smartness" the answer could be answerable.

Actually there isn't a strong link between neuron counts / brain size and smartness level, you can read something here. Einstein had a quite normal brain size, nevetheless he was without any doubt a genius and, for example, whales have a brain way bigger than humans. Because of the doubtful definiton of "smartness" also the processing speed could be relevant but not the prime requirement. The A.I. will be surely faster, which means that it could for sure process info faster than us, but it doesn't really mean that it will be smarter.

Try to imagine an iper fast ant, the concept is equivalent also for less smart entities. Would a i-f ant be more intelligent than a regular ant? No, it's simply faster and, even if the i-f ant will probably outperform the regular ant, it doesn't mean that the i-f ant will understand the beauty of Maxwell's equation or that it will develop sarcasm.

As we just saw intelligence doesn't strictly depends on brain size and brain speed, in fact our intelligence is mostly relying on brain "quality". It's more relevant the number of neuron connections/ramifications rather than their sheer number. We also have some specialized part of our brain, an these part are really valuable: it's their presence that make us "smart" and able to react to a wide range of situation (which could be a nice definition of smartness), not our brain size or speed. A twice as big and/or fast basal ganglia (reward and punishment feedback) but without a hyppocampus (spatial memory and navigation) is not a good deal.

Another example: it's not a coincidence that the most "smart" beings are mammals, and the main reason is their development of prefontal cortex. Would you trade "planning complex cognitive behavior" with faster or bigger "consummatory behaviors such as eating, drinking, defecation, and copulation"? Of course, the faster the better, but the neurons' count and the signal's speed aren't enough to determine if an A.I. would be smarter than us.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.