How the machines work

Member lesson

Why models cannot count, spell or draw hands

Lesson 9 of 13 in How the machines work.

Why AI art struggles with hands · Vox
Lesson

Three famous weaknesses, one underlying cause. Counting, spelling and hands all require the model to be right about discrete structure, and a process that works by gradually resolving texture has no representation of discrete things. There is no variable in there holding the number five.

pixels inencodelatentsmall, and organiseddenoiseheredecodepixels outfine detail is lost on the way IN, not fumbled on the way outwhich is why small type never survives
Pixels compressed into a small latent space, worked on there, then decoded back to pixels.
Two carefully drawn hands raised side by side on a sheet of paper. The left has five fingers; the right has six, drawn just as confidently. A small magnifying loupe rests over the extra finger.
Both hands are drawn with equal confidence. Count the fingers, because the model will not.
Member

The rest of this is for members.

You have read the opening of this lesson. The rest of it comes with a membership.

One membership covers all four courses and Brands.

  • All lessons in the four Learn courses
  • All brand breakdowns in the Brands section
  • All quizzes, flashcards and crosswords
  • An AI teacher on every lesson, for up to 60 questions a day
  • All course material published during the membership

Price

Membership is from $12 a month on the Yearly plan, which is charged at $144 once a year. Lifetime access is $59, paid once.Membership is from ₹200 a month on the Yearly plan, which is charged at ₹2,400 + GST once a year. Lifetime access is ₹3,999 + GST, paid once.

Check what stuck

Members only

You can't take this check yet.

Members answer these 2 questions and keep the ticket for this lesson. Every lesson has one.

  1. Question 1, locked

    Which of these words means this? “Getting the relationship between things right: three of them, behind that, holding this. Where image models are weakest.”

    • VAE
    • Glyph
    • Compositionality
    • Training
  2. Question 2, locked

    Which of these words means this? “Confident output that is wrong. In a language model it is a false fact; in an image model it is a sixth finger.”

    • Hallucination
    • Inference
    • Seed
    • Eval

Elsewhere in the library

Hear about new courses.

Leave an email and we will tell you when something new lands. We do not write often.