Asking people in an ice cream shop if they like ice cream is not a valid random sample because the results would likely be biased. Most of the answers would likely lean towards liking ice cream because most of the time, that would be the reason why people go to ice cream shops.