on boarding ChatGPT Image prompts

5 months ago

Create a South Park style cartoon of a fortified bomb shelter, walls made of concrete, with light and shadows highlights, but not dark, a thick, steel vault like door in the background. In this bunker sits an ornate table in the middle of the room, with three ornate chairs behind the table facing the viewer, on the right and left ends of the table are one ornate chair each, placed in profile. On the table is the board game RISK (image 1) lying flat on the table, set up and ready for play. In the left end chair sits Putin in profile (image 2), in the first chair facing the viewer sits Netanyahu (image 3), next to Netanyahu, facing the viewer is JD Vance (image 4) holding a tray of tacos, sitting next to Vance is Ayatollah Ali Khamenei (image 5) sitting on the right end of the table sits Donald Trump (image 6), a mix of snacks, such as pretzels, potato chips, and peanuts, and cans of Coke A Cola by each player. A lively bunch, excited as they play.

6 months ago

A wide image taken with a phone of a glass whiteboard, in a room overlooking the Bay Bridge. The field of view shows a woman writing, sporting a tshirt wiith a large OpenAI logo. The handwriting looks natural and a bit messy, and we see the photographer's reflection. The text reads: (left) "Transfer between Modalities: Suppose we directly model p(text, pixels, sound) [equation] with one big autoregressive transformer. Pros: * image generation augmented with vast world knowledge * next-level text rendering * native in-context learning * unified post-training stack Cons: * varying bit-rate across modalities * compute not adaptive" (Right) "Fixes: * model compressed representations * compose autoregressive prior with a powerful decoder" On the bottom right of the board, she draws a diagram: "tokens -> [transformer] -> [diffusion] -> pixels"