Look forward to your next visit!

Thank you, Andy, for your kind comment. I am back in Seattle right now, but I might visit China again the next year. I will let you know if I do!

I have been read your posts for a long time.
This post is again a great post.
When I read it, it let me come up with the time during my master degree.
As you wrote, at first I tried several ideas and failed.
But later, when I tried on another direction, some failed ideas I tried before come to my mind. Thanks to that idea, I finished my thesis.

I am so glad to hear that you are on vacation in China.
If you pass Beijing or stay in Beijing now, I would like to invite you to have a drink or a meal if you can.

I agree. It is true what you say. Quality of ideas will always more important than quantity. However, there is also strong evidence that the most creative people are those that have the most ideas. They try again and again until things work. When things work, these creative people are usually associated with the successful idea rather than the plentitude of unsuccessful ones. If you look at inventors this is very common. Scientists also fail very often but usually, their failed ideas are not known or published. All of this might make it seem that quality is more important than it really is.

I think one major reason for the innovation in deep learning is that you can test more ideas quickly if you have good hardware. So, if you are playful abd you have great hardware, you might be even more creative in that you can try more ideas in a shorter time. However, I would also say that “too much hardware” can stifle creativity. The application of lots of hardware might or might not indicate the degree of creativity of an idea, but if using more hardware stifles creativity probably is not a sign of lack of creativity but a sign of lacking other characteristics such as foresight and understanding what research is important, needed, and replicable by other researchers (that lack the hardware that oneself has). A stereotypical example of using too much hardware unwisely is when you use a lot of hardware for the sake of using a lot of hardware, for example, NVIDIA’s super-large transformer would fall into this category. However, onr could imagine that lots of compute is used for a creative project. For example, the early Google brain project to do unsupervised learning on a large collection of youtube videos could be seen as a creative way to gain deeper insights into how features relate to data.