nottombrown

Author here: Sorry for the confusing formatting on the task descriptions at the end of the paper. That "4" is the human-generated target completion, not a model generated completion. I'm not sure whether the model got that particular question correct, but from Table 3.7 that GPT-3 has 36.5% accuracy on DROP in the few-shot setting.

Many other readers were confused by this so we'll update the formatting to say "target completion" to make this more clear.

I agree that the phrase "non-specialized" is a bit tricky. Ants are mostly non-specialized in structure (there are only a handful of forms within the colony), but individuals take on specialized behavior based on the conditions that their put in. I think stem cells could be a better metaphor.

Nobody has mentioned hooktheory.com yet. If you're interested in why some music sounds good, check it out.

This is great. I could see it replacing much of my Illustrator work.

Thanks for bringing this to the playa, Samuel. Hope to see it again next year!

In Swift, you have to signal and guard against the nil.

In ruby, you choose whether to signal and you choose whether to guard. Here are some ruby examples that may be more clear:

Does not signal

    user = User.find_by_favorite_dinosaur("Dromiceiomimus")

Does not guard

    def print_favorite_dinosaur(user)
        p "User's favorite dinosaur is #{user.dino})"
    end

Signals and guards

    def print_favorite_dinosaur(user)
        if user
            p "User's favorite dinosaur is #{user.dino})"
        else
            p "No user. (Let's assume she likes T-rexs)"
        end
    end

    possible_user = User.find_by_favorite_dinosaur("Dromiceiomimus")

    print_favorite_dinosaur(possible_user)

Author here. Let me know your thoughts, concerns, and favorite dinosaur.