> When evolution invents a chemical it's completely random and untested.
This is where I think your argument breaks down a bit. Let's take your example of sugar: humans have been consuming sugar since before it would be reasonable to call them humans. That's why our bodies have adapted to crave sugars and seek them out--because they have been tested by evolution and found to contribute to survival and reproduction.
And when you say that sugar is decidedly toxic at average consumption levels, you're correct, but that's a human-created problem: the levels of sugar naturally occurring in the fruits a person could obtain, say, 10,000 years ago, were much, much lower than the levels we can obtain by harvesting massive fields of sugar cane, sugar beets, and corn, and refining the sugar from them.
So evolution did test sugars, and then humans refined them to a far greater extent than nature did and, without adequate testing, integrated sugars into our diet to an extent that's culturally very difficult to reverse even though we now know it's deadly.
And the same goes for a lot of things: coca, consumed in leaf form, is relatively safe and effective for treating altitude sickness, but when refined and mass produced becomes cocaine. Alcohol when consumed from rotting fruit has little danger until it's produced in quantity and consistency that enables alcoholism. Chances are, if there's a whole chunk of plant that humans have been interacting with for millennia, it's pretty safe because it has been tested, but refinement processes invented by humans bring otherwise safe chemicals into concentrations and quantities that become unsafe. And while we do test these things, it's often after-the-fact, too late, or just completely ignored.
To be clear, I'm not advocating a "things from nature are better" stance. GMO grain has saved millions of lives. Most of the medicines out there are the results of human innovation. And there are plenty of poisonous plants out there which are perfectly natural.
What I am advocating is that the human track record versus evolution track record is pretty bogus, and vulnerable to selection bias. The fact is, that's just not even a good way to choose whether to consume something. We rarely have enough information to make an informed choice. Even when we study something for a decade we don't completely understand that. So in the absence of information, basing our decision on "organisms pretty biologically similar to me have survived by interacting with this other organism for millions of years" is not a metric to be dicarded.