Deep Learning Deep Dive

The actual paper for DALL-E was released only a few days after we published our episode 2, so in this episode we re-visit DALL-E in its full published glory. Joining us as a special guest for this episode is Aditya Ramesh from OpenAI, the lead author of DALL-E. DALL-E blog post: https://openai.com/blog/dall-e/ DALL-E paper: https://arxiv.org/abs/2102.12092 DALL-E code (encoder/decoder model only, so far): https://github.com/openai/dall-e   Deep Learning Deep Dive is also available on YouTube, where we scroll through relevant parts of the paper and code while talking about them: https://www.youtube.com/watch?v=PtdpWC7Sr98 We reached out and collected written consent from all participating audience speakers.

Andrej Karpathy and Justin Johnson deep dive into OpenAI's DALL-E and use it as an anchor point to recurse into some of the recent work in AI on image generation. Approximate agenda: DALL-E Blog Post: https://openai.com/blog/dall-e/ ImageGPT https://openai.com/blog/image-gpt/ VQ-VAE https://arxiv.org/abs/1711.00937 VQ-VAE-2 https://arxiv.org/abs/1906.00446 Gumbel-Softmax / Concrete Distribution https://arxiv.org/abs/1611.01144 https://arxiv.org/abs/1611.00712 VQGAN https://arxiv.org/abs/2012.09841 Andrej's attempted re-implementation of VQVAE and GumbelSoftmax: https://github.com/karpathy/deep-vector-quantization/blob/main/model.py You can see a video version of this episode on YouTube: https://www.youtube.com/watch?v=gMc90bqHMSM We reached out to all speakers and obtained their written consent to appear in this recording.