He meant about this matter. He is also wrong.
He is wrong because he makes the mistake that a finding of no statistically significant result means the result doesn't provide evidence.
To see this is wrong imagine you initially thought people were probably (but not certainly) likely to react to these pictures in a racist way. Scientists perform larger and larger surveys never surveying every person but failing to find any statistically significant difference with arbitrarily large populations (assume for simplicity there is an arbitrarily large number of humans). No matter how likely you found the question initially at some point you will find the result so improbable if the claim is true that it provides enough evidence to reject the claim.
Statistical significance is a trick for giving a gauge of how persuasive you should find the result given your priors. Since different people have different priors it's not usually useful to assume a certain prior probability distribution of results, e.g., the probability that people are at least X% more likely to judge a black woman negatively when breastfeeding. So we tell people the significance of a study and if they want to know how it affects their beliefs they figure out just how surprising a result of that kind with that level of statistical significance is on their model if the claim is true and if it is false. Theoretically they could apply bayes theorem to that...in practice we use a more heuristic approach. To see this has to be true note that no study will shift your belief if you think it is already true with probability 1 or 0.
Then again this is almost impossible to teach in a full year stat course so few people who don't already know this are likely to understand.