One problem we have at work right now is putting models into production. Putting one model into production is not super difficult (not easy, but can be handled by brute force), but getting to a point where you have 10+ models in production is quite overwhelming to handle. Do you have any tips on managing a large number of models (tracking them, putting them into a system, building a review process around them)?
Do you have an opinion on how likely it is for quantum computing to have a practical impact in ML in the short term (say, next 10 years)?
Hello,
Great to see that initiative.
I know you had the change to assist last week to NeurIPS. Would you mind share your personal choices on interesting ideas/papers that were shared in the conference?
Thank you in advance
Now that MLPerf is out, do you think fastai will try their hand at the open division? (allows anything between dataset and trained result - so any new algorithm ideas are allowed).
What will replace Deep Learning? Deep learning could have not been adopted until we got large amounts of data and large computing power. Is there a technology now that may benefit from developments in the future? Loved the class! Can’t wait for second part.
With neural nets staring to perform better and better with general tabular data as well, should we still focus on traditional ML algorithms? (apart from Random Forest and Decision Trees which are helpful for determining feature importance)
If yes, then which ones?
What new things are you learning right now and what is your process of acquiring a new skill?
Great course, thanks Jeremy! My question is: the second part of the course will be available as this one? I mean will be open to the public from differents parts of the world?
fastai DL courses related: for someone planning to take part 2 in 2019, what would you recommend doing/learning/practicing until the part 2 course starts?
Once there was a period of “AI Winter” when people become disappointed about Neural Nets (and AI in general, I guess). How do you think, is it possible to get into a similar situation with modern AI techniques in near future? Or we’re getting into a stage of unbounded growth of new methods, architectures, etc.?
What do you think is to happen to news vs fake news with the advances in deep learning fakes for videos and voice? Are you worried?
I haven’t used their service but based on the
videos I have watched, lyrebird.ai does exactly that, train a language model on very limited data.
This needs more likes. Lol.
Where do you see the fastai librairy going in the future ? Say in 5 years ?
What would be the focus of fast.ai DL1/DL2/ML courses in the coming 2-3years? Any books to be published as you have said in one of the podcasts? Any course on Deep Reinforcement Learning in the near future?
What’s your take on Quantum computing?
What are the shortcomings of the fastai librairy right now ? How can we help ?
What is the best technique to compress model for hosting in web?
Deep Learning For Coders - Part 2 Question: Do we know if part 2 fellowship will also be available for global participants? This has helped a lot of folks globally!
How do you think, is it possible for one without an AI/ML degree from the well-recognized university to get working into a company like OpenAI? Do you think to work on your own projects, portfolio, and competitions is enough to get a good Deep Learning and/or scientific skills?