AI will be megalomaniacal

I disagreed with Lubos Motl about unification, and I also disagree with his view of AI:

So even if these machines achieve high intelligence, there's no reason to think that these machines will have megalomaniac goals! Unless someone "programs them" with the goal of doing something bad to the whole world – and in that case, the human-creator is the main entity who should be held accountable – the robots just won't get obsessed with megalomaniacal goals by themselves because they haven't gone through the evolution process that could train them to become power-thirsty or megalomaniacal.

Facebook today is a giant megalomaniacal AI system that has been programmed by a megalomaniac, Zuckerberg. Likewise with Google, Apple, and Amazon.

I must point out that leftists have been saying similar things for more than a century – especially the statement of the form "the free market must already surely fail in this new human activity, that one etc." – to defend the establishment of some form of communism or totalitarianism in a section of the human activity. They were always wrong. The free market doesn't break down when it's applied to newspapers, radios, TVs, videos, songs, books, computers, computer programs, telecommunication, mass transportation, water pipelines, and lots of other things. All the words they have ever claimed to be arguments were just illogical piles of nonsense and in all the sufficiently old disputes of this kind, the leftists have been proven wrong. There exists absolutely no reason to think that the case of the AI is any different.

When there are 100s of millions of customers, there are network effects that result in a winner-take-all economy.

When ppl complain about bias in Google searches or iphones killing popular features, the companies just imply that you are too stupid to know what you want.

Yes, I know that there are other search engines and phone makers. I use the alternatives myself.

The threat is that large integrated systems will have an intelligence of their own, and ppl will become dependent. Already there are drivers who cannot find where they are going unless they submit to dictatorial orders from Google. In the future, ppl may be submitting to orders on a wide range of matters. Maybe the robots won't get obsessed with megalomaniacal goals by themselves, but they will get megalomaniacal anyway.

Perhaps Motl would argue that market forces will result in programmers constraining the AI systems to be less megalomaniacal. Yes, Facebook would be worse if it were not for some occasional user protests. But in my opinion, it is pretty bad and is going to get worse.