People take educated risks all the time, we do not calculate (and sometimes can't) every possible outcome of actions, saying we should stand still to avoid anything going wrong is just regressive at best. If you follow that train of thought you should use no modern medicine, the complexity of a living being is too high. We should and can take heavily educated guesses to improve our state of being.
In highly complex systems, educated guesses are no better than random picks.
You just have to be aware what kind of system you're dealing with, and be prepared to deal with the consequences. Also, keep in mind time horizons – very important.
That's nonsense. Regardless of complexity of the system, a single bit of knowledge in your educated guess cuts the space of possible choices in half. Even a few bits of knowledge can let you avoid dangers that random selection will reliably hit.
Yes, but most decisions don't intentionally make permanent heritable changes to a species. It's not convincing to say the worst-case scenario associated with GMO monocultures is the same risk associated with, say, a person taking a new drug. One could only effect one person, and that person's consent is possible to obtain. The other could effect millions of people, and is almost impossible to obtain their consent. (Arguably, the anti-labeling campaign is an attempt to make getting such consent irrelevant).
Gosh, just read his first book. Black Swan. Where he shows that before Europeans discovered black swans in Australia it was a perfect "educated guess" to assume all swans are white.
Welp, Taleb's argument is "scientism", pure and simple.
A lot of interesting problems in real world seem NP-complete, and we deal with them just fine. NP-complete doesn't mean "any wrong step can kill you"; it means we need to be satisfied with almost, but not quite, optimal solutions.