As it's currently understood, the information that AI will file, analyze and use to make decisions will be scoured from the internet. The verbal ideas and opinions of Cambodian farmers, Canadian truck drivers and Colombian cocaine traffickers won't be grist for the AI mill. Probably the material most desired for data center storage will be gleaned from things like this, containing ample statistics that can't be verified as guidance for future activity.
In fact, the only purpose for extensive files of any information is to guide later decision making. Gibbon's studies of the decline and fall of the Roman Empire were meant not just for amusement but also to point out the failures of that nation perhaps in the hope that they wouldn't be repeated. Judicial records of a person's behavior can determine his entire future. Before today the positive or negative effect of decisions made to achieve a certain outcome had to be measured against the information used to make those decisions. In the future, the total information that AI uses to make choices will be too complex for human analysis. AI systems will need to be used to validate the findings of previous efforts in a never-ending loop.
Evidently an AI conclusion, should one be reached, will supersede that of any lowly human or all this investment is in vain. Thus if AI says drop the bomb on the Kremlin, that will be the thing to do. No level of "natural intelligence" will be sufficient to negate the AI findings. Who will have access to this AI "opinion" is another question.
Man may be capable of doing many things but considerations must be taken into account. The findings of AI, whatever they may be, are incredibly expensive. Giant data centers, filled with esoteric semi-conductors, filing and rearranging information, are using humongous quantities of power that's already putting a strain on the national electrical grid. What's more important, a town washing its clothes and microwaving its Eggos or a data center making an informational omelet? This isn't a new issue. The Amish made their choice many years ago.
tripsavvy.com
In the Jan 4-5 issue of the Wall Street Journal Ben Cohen describes the work of ASML engineer Breanna Hall who looks after an Extreme Ultraviolet Light lithography machine in the Boise, Idaho fab of Micron Technology. He describes it as "the most indispensable machine in the world." This is because the semiconductor chips it manufactures are used in so many high tech products, appliances, electronics equipment, automobiles, etc. The new EUV chips are simply extremely small versions of earlier semiconductors. The fact is that most of the products that now contain semiconductors existed before they became available and the world got along fairly well. A new ultraviolet lithography machine costs as much as $370 million. There might be an increase in productivity in the use of this machine but productivity is figured in dollars and hours. Is that all there is in life? If the EUV machines disappeared from the face of the earth corn would still grow, cows would continue to produce milk, crude oil would still be pumped from the depths.
It could be the case that drivers need and want "value added" features in automobiles like GPS systems or self-driving capabilities but for well over one hundred years they puttered from point A to point B without sophisticated and unrepairable components that now make up most of the cost of a car. Isn't there a point where advancing technology simply isn't worth the price?
earthtrekkers.com
Moderns have always wondered about the construction of incredible projects of the past, the pyramids of Egypt and the similar structures of MesoAmerica, the fantastic amount of labor and engineering that went into what now seems to be a waste. How could it have been possible to coerce people into devoting so much energy into what must have seemed very important to someone else? The AI data centers may inspire similar questions from future humans, should they come to exist.