The first thing that most people don't understand about "artificial" intelligence, is "organic" intelligence. Because very few people are familiar with quantifying human experiences, there is no wide-spread and commonly agreed on state of this art.
On the other hand, the venture capital marketing cycle has popularised wide-spread and commonly agreed on states of the art, for machine learning as of 2024. Unfortunately these are commercial distractions from fundamentals, which will have their time, and pass. They will have their time now because many of these currently popular concepts have had their time before, being hardly new ... their predecessors sprung up in previous summers and wilted before winter, time and time again. What has changed of late is mainly a matter of decreased costs of computing - the algorithms didn't get much better, but it got a lot cheaper to run bad algorithms. There is a place in the world for things like neural networks, but they are not the main hurdle for building machines which are indistinguishable from meaty humans.
Know thyself, it is said. Human consciousness consists of specific data structures. At the "micro" level, the atoms of experience are merely functions in fields of multi-modal sensation. At the "macro" level, the chemistry of social interaction is largely down to classifications of motivation and emotion.
Back to commercial viability. What can we look forward to? Well, most superficial art and design problems are down to the "microaesthetic" sort of data structure. Whereas, ( sales and marketing ( is talent management ( is politics ) ) ) is pretty much down to "macroaesthetic" data structures.
None of this is really new to me - since I worked on it in college a couple of decades ago. I just always when the rest of the world will figure it out.
Meanwhile, this week I have been working on my own physical and mental conditioning, in a transitory phase between projects. I have been working on forgetting things faster, so that I can reuse my short- and mid-term memory for new things. As a heuristic, I reminded myself today, that I should think of my own short/mid-term memory management as a matter of "building models" and "destroying models". Models are simply sets of vertexes between data. The same sorts of data I referred to above.
No comments :
Post a Comment