 So, let's briefly talk about what we learned today. We learned about classical confnets, the kind of confnets that started revolutionizing image recognition and that solved ImageNet. We learned how Alex networks and how it's really amazing engineering that's behind it. We learned how, in many cases, deeper networks win, but also how skip connections can massively help in that endeavor. We learned how confnets are broadly used and how there's a lot of prompts in computer vision where they are really state of the art and useful. And then we, but the upshot here is, in a way, object recognition for the kind of task that ImageNet is about, feel solved. How is it solved? Well, we have these massive data sets and we have these really, really big confnets that are able to solve these tasks in a lot of cases better than humans. But keep in mind, the humans understand something deeper about it. Say we understand the essence of a dark and so we will be able to recognize darks if the images are very different. So humans would be much better at recognizing drawn darks or darks in space and things like that because they have the ability to think beyond the simple confnets like feature detectors that we have. They understand something about the essence of darks. Now, make sure you give us feedback. We're now halfway through the course. You helped us make the course better in so many ways. I hope you realize that we read all your feedback and we incorporate it everywhere in the course whenever we can. And keep up the great work. Give us feedback. Thanks for participating.