Deep learning is a highly effective set of techniques within AI that utilizes neural networks to learn from data to solve specific problems.

“What's been most uh very effective in the recent about 15 years is a set of techniques that fall under the flag of deep learning that utilize neural networks. It's a network of these little basic computational units called neurons, artificial neurons.”

Deep learning is a foundational technology that allows for extracting complex information from data, such as identifying objects or text in images.

“instead, what what you do is you you you put these what they're called neural networks. They look at all of the pixels of the image and and you give it lots and lots of examples of images that have somewhere it's got the text in it and you tell the neural network what the text is.”

Building models capable of aligning with classical algorithmic computation is essential to overcome current shortcomings in reasoning and generalization.

“I still believe that like building machine learning models that are capable to align to classical computation is going to be really really important to address the shortcomings that are not so easily plugged by just gathering a better data set.”

Deep learning models do not inherently overfit because they perform layer-wise compression and denoising that effectively restricts the solution to low-dimensional structures.

“Because if the neural networks is try to the operator is try to compress, try to realize certain contracting map, compress volume, they will never overfit, right? Even I over parameterize, it will never overfit.”

Deep learning is fundamentally limited because it prioritizes likelihood over truth and lacks the ability to guarantee correctness.

“this thing about like um not knowing the truth or not knowing basically the difference with truth truth and likelihood is a fundamental issue”