The 42nd episode of this podcast is about the technology that is currently seen as the answer to everything and nothing: ChatGPT.
Andreas Köpf from Open Assistant is visiting Nico Kreiling this time. The two discuss how ChatGPT kick-started the technology revolution and what implications it has for the future. The conversation first goes back to the basics of ChatGPT. It takes an in-depth look at the Reinforcement Learning from Human Feedback (RLHF) that helps Large Language Models follow users’ intentions.
The post introduces Open Assistant as an open-source alternative to ChatGPT, which has built a dedicated community that has released the dataset and models with a free license. Nico and Andreas discuss which hardware and models were used for the training. Ethics is an important issue because when creating the training data, you have to be careful about the correct answers to potentially dangerous questions.
Finally, the discussion partners take a look at the social dimension and discuss the upcoming transformations and next development steps.
Recommended Editorial Content
External content will be loaded here with your consent.
(rme)