How did DALL-E teach AI to imagine pictures from words?
Type 'an armchair shaped like an avocado' and watch an AI paint it from scratch.
The answer
Announced by OpenAI in January 2021, DALL-E was a 12-billion-parameter version of GPT-3 modified to generate images from text descriptions. It worked by encoding both text and images as sequences of tokens and training a Transformer to model them autoregressively. DALL-E could combine unrelated concepts ('a snail made of a harp'), apply styles, and even render text within images. Its successor, DALL-E 2, released in April 2022, used a diffusion model approach that produced far more photorealistic results at higher resolution.
Source: OpenAI
- dall-e
- image generation
- openai
- multimodal
More from AI & Machine Learning
- What radical question did Alan Turing ask in 1950 that launched the entire field of artificial intelligence?
- Can a machine fool you into thinking it's human - and does that mean it's truly intelligent?
- What happened at a small Dartmouth workshop in 1956 that gave artificial intelligence its name?
- How did a 1958 invention called the Perceptron convince people that machines could learn like brains?
- Could a 1966 chatbot really make people believe they were talking to a therapist?
- What was the first robot that could 'reason' about its own actions?