There are two points being made here by you and the original commenter:
1. Social interactions at commercial centres are valuable and Amazon and others who follow will be unable to offer such value.
2. Social interactions at commercial centres are more valuable than the beneficiaries are conscious of.
#1 I think we all agree on. #2 is the critical point and the degree of which is arguable and would swing Amazon's innovations between good and bad.
Let's imagine that on average, whenever someone goes into a shop and interacts with another human they are gaining 50% of their social wellbeing (i.e. a lot). Let's also imagine that on average whenever someone goes into a shop and interacts with a human they leave believing they have gained only 1% of their social wellbeing (i.e. not much). Such a person is likely to stop shopping at a shop with tills and cashiers in favour of a humanless shop such as Amazon Go. However that would likely be a mistake because they would lose half of their social wellbeing without even realising. This in my opinion would put automated shops into the category of a socially 'bad' thing.
Now imagine the actual social benefit of traditional shops is 5% (not huge but not insignificant) and the perceived social benefit is also 5% (i.e. we're fully aware of this benefit). In that case, people should be able to make the choice of which kind of shop is best for them based on a complete set of information. This in my opinion would put automated shops into the category of a socially 'good' thing. They offer more choice coupled with the relevant information to make that choice.
The discrepancy of perceived vs actual benefits can explain why we make the wrong choices with regards to a lot of market decisions from driving a gasoline car to eating at a restaurant. The best way to determine whether a market option is socially good or bad is to try to measure this discrepancy, see if it exists and if so, is it significantly large?