Be sure to check out part 1 for context.

• These days, AI means LLMs plus diffusion models.

  • LLM means "Large Language Model" because its trained on so much text

  • Diffusion models train computers to turn random noise into images.

  • Since each image starts with different random noise, the same prompt can yield different images.

• We can create new images from existing images and a prompt to change it. Or we can create an entirely new images with just a prompt.

• AI generated images look passable at first glance, but have flaws when inspected.

  • Letters blur or are misspelled.

  • Logos and symbols aren't exactly right.

  • Shadows are pointing the wrong way.

  • The textures look off.

  • Fantastical situations.

• Anyone can use these image generators for good or for nefarious purposes. That's why it's important to learn how to spot AI generated images.