Interview with Juergen Schmidhuber at Data Christmas 2020

02:00-05:38 What do you think were the most outstanding underestimated news and achievements in AI field in 2020? 05:41-11:28 What do you think about trends in ML like transformers trying to replace LSTMs in NLP? 11:29-16:06 Are you working on any new types of models right now? 16:07-20:41 What is your opinion on the most underestimated ML subfield like Reinforcement Learning? 20:42-22:17 Your best recommendation for our community is to look into AI in the real physical world, right? 22:18-33:10 Do you think it is possible to achieve great results in creative AI, particularly in subjective beauty? 33:17-35:50 What prevents chat bots from reaching more intelligent levels? 36:03-39:39 What is your opinion on broader impact statement at neurips, should it be applied to neutral topics? 39:40-40:38 Hopefully we will see less examples of prohibition of some articles because of their statements in the future. How would you describe such statement for your LSTM paper? 23 years ago, could you have guessed what would be its potential uses and misuses? 41:12-48:18 Could it be dangerous or risky that world’s leaders might have not clear understanding of the modern AI and its capabilities? Can politicians ban experiments with GPT6 or GPT7 technologies, for example? 48:31-50:37 Do you think it is possible to construct an error function for AGI so that it would find an optimal behaviour to help people without taking their jobs? 50:54-53:29 What do you think of the possibility of AI winters or similar e-facts in the future? 53:48-56:48 Where can we buy the hat that you wore on most of your pictures? 56:52-58-03 What is your recipe for staying in a good shape? 58:26-59:00 If you could send a message to yourself when you were 18-years old, what would it be? 59:03-50:15 What is your Christmas wish for the AI field in 2021?