The whole point of inventing programming languages was to make it possible for less skilled people to write code. When you had to write computer code in binary, hexadecimal or assembler, you had to be really, really good. Most people couldn't do it.
Programming languages removed the need to think in binary or hexadecimal. You could think using English language equivalents instead. This made programming a lot easier, thus allowing lower IQ people to code. But, allowing stupider people to code meant you not only vastly increased the number of software coders, you also vastly increased the number of defective programs. This is how software turned into the security nightmare it is today.
When AI creates computer code, it produces the statistical average of all the code out there. AI removes the bottom 50% of coders from the job. Given the Pareto principle (80% of consequences come from 20% of causes), the bottom 20% of coders were probably causing 80% of the security and integration problems.
Since AI removes the bottom 50% of human coders, having AI produce code means most of our software security and integration problems will go away. This is likewise true in every skill area where AI is applied.
Is AI perfect? Of course not. It makes stupid mistakes because it trains on human-generated data, which is filled with stupid human-generated mistakes. AI is just a mirror of human activity. It is us watching the average human intellect doing algorithmic work. But, once you fix an AI algorithm or a data training set, it stays fixed. Software doesn't degrade. 2+2 = 4 for all eternity, it is an equation outside of time, and that is true of every algorithm, whether AI uses it or not.
So, AI starts at the statistical mean of every data set. It starts out being roughly as good as the average person doing the work. That means it starts out already better than half the people doing the work.
Once you remove the bottom 50% of error-ridden data and the bottom 50% of corrupted algorithms, AI is now at the 75th percentile, not the 50th percentile. Keep iterating that process and AI quickly becomes "brilliant". That doesn't happen because AI thinks, it happens because we painstakingly work through the algorithms and the data sets, stripping out the errors and leaving only the good algorithms and data behind for AI to use and continue to train on.
The big complaint about "science" in the last forty years is precisely that so much of it is pure crap. Endless examples of peer-reviewed published articles that aren't worth the paper they are printed on because both the researchers and the "peers" who reviewed the articles were below average in competency (and 50% of any human population is, by definition, below average).
AI solves the problem by stripping out the bottom half. Yes, it absolutely still makes errors, but it makes a lot fewer errors than the bottom half of the human population makes. Once it is optimized, it makes fewer errors than the bottom 80%, 90%, 99%.
It used to take decades to train people to master complex tasks, master those tasks so well that even when the trained person is hung over, exhausted, sick, or otherwise incapacitated, they could still produce the necessary result. Once an AI has algorithms and data sets optimized, that information is infinitely replicable, it can now be copy-pasted into an infinite number of machines. Given a sufficiently nimble robot, i.e., a machine that can effectively interact with the environment, that means we don't have to spend decades training individual experts in a bespoke process. We can copy-paste and we have an infinite number of experts.
The power of AI is not just in its promise of comparatively error-free operation, i.e., compared to humans, but in the instantaneous and infinite replicability of all knowledge and every technique. Teachers often crow that they "teach people how to learn". It was never really true, but now, for at least the bottom half of the human intellects in the world, it is not even necessary. Algorithms and data are now built on an assembly-line. We get the results without the work. Work literally disappears.
In Judaism, Christianity and Islam, even God has to work in the clay of the earth to build up a human body. The AI-robotic infinite replicability of technique and knowledge means the very foundational understanding of the universe that Abrahmic religions provide will no longer correspond to the world we experience. We will all become Deists, hands-off spectators watching the tools we made produce the results we need in a clockwork universe.
Christians believe God eternally breathes forth the single Word through which all creation comes into being: Father breathes forth Spirit and the Word. With robotic AI, we will need but to speak, and lo!, it will be made. Like God, we also will speak reality into existence, we will watch the clockwork production of our vision being built before us. Even the stupidest among us will be able to do it. Whether we want this or not, that is what we are building.
No comments:
Post a Comment