In this episode, I am with Aaron Gokaslan, computer vision researcher, AI Resident at Facebook AI Research. Aaron is the author of OpenGPT-2, a parallel NLP model to the most discussed version that OpenAI decided not to release because too accurate to be published.
Better Language Models GPT-2
In this episode we discuss about image-to-image translation and the dangers of the GPT-2 model. As a matter of fact there is no better language model than GPT-2. With the last version of GPT-2 it is possible to generate paragraphs of text, given a seed. Such paragraphs actually make sense and it is quite challenging to understand if the author was a human or an algorithm. Therefore, anyone could use GPT-2 to generate fake news or facts that are not supported by any evidence.
Better Language Models GPT-2
In addition to this, we speak about the future of Artificial Intelligence. I asked what would a researcher like Aaron see happening in the next 5 years. Aaron believes that self-supervision will play a fundamental role in the future of AI. Moreover, he provided us with some very interesting resources and demos that will definitely blow your mind!
Enjoy the show!
Join the discussion on our Discord server
References
Multimodal image to image translation (not all mentioned in the podcast but recommended by Aaron)
Pix2Pix:
https://phillipi.github.io/pix2pix/
CycleGAN:
https://junyanz.github.io/CycleGAN/
GANimorph
Paper: https://arxiv.org/abs/1808.04325
Code: https://github.com/brownvc/ganimorph
UNIT:https://arxiv.org/abs/1703.00848
MUNIT:https://github.com/NVlabs/MUNIT
DRIT: https://github.com/HsinYingLee/DRIT
GPT-2 and related
Try OpenAI’s GPT-2: https://talktotransformer.com/
Blogpost: https://blog.usejournal.com/opengpt-2-we-replicated-gpt-2-because-you-can-too-45e34e6d36dc
The Original Transformer Paper: https://arxiv.org/abs/1706.03762
Grover: The FakeNews generator and detector: https://rowanzellers.com/grover/
