Neural networks can struggle with images they haven't been trained on, highlighting the continued importance of handcrafted models in certain contexts.

The introduction of large neural networks has allowed us to solve complex problems in vision, speech, and language, shifting focus from traditional handcrafted approaches to machine learning.

The integration of handcrafted models with neural networks presents exciting opportunities for mathematicians to explore and improve image denoising techniques.

Understanding the properties of handcrafted models can help inform the development of neural networks, particularly in terms of stability and performance.

Neural architecture search allows us to generate models automatically, which can outperform human-designed models in specific tasks without human intervention.

While handcrafted models for image denoising are still relevant, deep neural networks are increasingly outperforming them in various scenarios.

Our group investigates the hypothesis that large amounts of compute can solve interesting problems using neural networks, which has led to significant advancements in unsupervised learning at scale.

Deep learning and neural networks are shifting how we approach many problems, making neural networks the best solution for a growing number of challenges.

Currently, we train bespoke models for each problem, but the goal is to develop models that can generate solutions for multiple problems, enhancing architecture search.