Hello Deep Learning • Bert Hubert • GOTO 2023

Join Bert Hubert's in-depth exploration of the deep learning process, debunking myths and sharing practical insights on understanding and utilizing neural networks, from recognizing patterns to cautioning against overreliance on automation.

Key takeaways

The process of training a neural network resembles playing with a black box, where you tweak parameters until it works.
The network doesn’t know what it’s doing, it’s just a random matrix in the middle.
You should be skeptical of deep learning, but also experimenting with it can be fun.
There is no magic, it’s just people working on it and making progress.
You should want to understand how it works, not just use the libraries without understanding.
TensorFlow and other libraries are just tools, not magic.
The process of deep learning is not unlike playing a game, where you tweak parameters until it starts working.
You should not be afraid of AI replacing jobs, it’s more about augmenting human capabilities.
Generative AI can do amazing things, but it’s still a developing field.
The author is worried about the potential negative consequences of AI, such as losing autonomy.
Networks can recognize patterns, but they don’t understand what they’re recognizing.
You can use deep learning for tasks like recognizing digits, and it can be impressive.
The author doesn’t think deep learning is a big deal, but it’s still an important field.
You should focus on understanding how deep learning works, rather than just using it.
There are manyuse cases for deep learning, but it’s not a magic solution for everything.
The author is skeptical of AI replacing humans, and thinks it’s more about augmenting human capabilities.

Hello Deep Learning • Bert Hubert • GOTO 2023

More talks