I have been following Hoffman's work for many years, thanks to a mutual friend.
A fascinating talk (I admire good public speakers), but I disagree with his conclusions. Evolution does give accurate perceptions, but only accurate enough. Evolution is rarely wasteful. The example of the beetle getting confused about bottles was a bad one, because evolution had not been allowed to act. Given a few thousand years, the beetle would almost certainly have evolved to distinguish between female and beer bottle. If you are going to assess evolution's power to improve accuracy you have to wait for an evolutionary time scale. The beetle's vision isn't telling it where females are - after all, beetles have no concept of females! If you assume that this was what the beetle's vision was trying to do then of course you will come to the assumption that it is wildly inaccurate. The beetle's vision was only showing is what it would have to see to mate, and that had been very accurate for millions of years. The appropriate measure of accuracy is between what we experience and what we believe we experience.
We have pretty accurate vision. The tomato really is in front of us. We throw that because we can objectively measure how far a hand has to move to reach it. Apes can accurately assess what other apes can see so that they can hide food. That hiding involves objectively accurate assessments of lines of sight.
Of course, our vision is only accurate to a limit. We can't see the tomato's cells or its quarks. But that doesn't mean we are mistaken in chopping it up and putting it in sandwiches.