|AI primer - please feel free to add to or correct this explanation.|
Posted by: LateForLunch ® |
Author Profile Mail author Edit
The recent increase in discussion of AI, mostly centered around GAI (General Artificial Intelligence) AI Chatting brings up the whole issue. One might wonder, why now? Where has AI been all of the time before this?
The answer is that AI underwent a surge development in 2010 after a couple of false starts.
There was focus on LISP (list processing) programming but after a strong start that path eventually collapsed Then more attempts with other sorts of programming and hardware. Finally after some powerful software and hardware were developed it allowed significant progress in the field.
Internet search engines used some of that new AI technology to process information about preferences with good success financially. There were some bad effects also in terms of the AI elements. 'Not necessary to go into that in detail now.
So they found a way to simulate organic neurons using advanced hardware and software. The human mind does "calculations" which are not linear like a machine using binary code - human minds use associative networks. When
human brain tries to do a calculation, the data travels along a
chemical-electrical pathway in which many neurons peripheral to the main path also add to the calculation. Memories of similar questions, pertinent information that involve emotional impressions, hunches, creative elements, eventually contribute to the calculation. In a linear network, all that happens is a stream of data mixes with other streams of linear data so all that happens is a vast, adding/subtracting problem that must be translated into some subjective conclusion to be useful.
What the technology has succeeded in doing is simulating an associative network using a linear computational process (addition/subtraction equations) in which millions of artificial neurons contribute to the calculation.
The human mind does about 20 billion FLOPS in an average computation. Thanks to increases in the ability of machines to perform far more than 20 billion FLOPS, combined with software that allows linear binary-language-based machines to perform a similar associative calculations, sophisticated functions previously beyond the scope of machines have become possible.
Modified by LateForLunch at Wed, Jul 26, 2023, 00:39:53
|Post Reply | Recommend | Alert||View All||Previous | Next | Current page|
|Replies to this message|