Alright, so far DiscoverArtificialIntelligence has been more about talking about experiences relating to AI, not truly showing you some of the new discoveries. It’s important that we discuss some of the more recent breakthroughs, as that may be more convincing and interesting information.
To start off, here is a quote by an article that I found particularly interesting:
“The first, a vision network, analyzes an image from the robot’s camera to determine the location of objects in reality (in OpenAI’s video example, these objects are blocks of wood on a table). The network is able to do this despite never having seen the actual table or blocks before. Instead, the researchers trained it using hundreds of thousands of simulated images, each featuring various permutations of lighting, textures, and objects.
The second, an imitation network, determines the intent of a task it observes a human demonstrating via a virtual simulation. It then imitates the task in the real-world setting. Again, this network was trained on thousands of virtual demonstrations, but none that took place in reality.”
What I truly love about this article is about how its able to convey the importance of the technique that they are using in a simple and thoughtful manner. Essentially, this vision network is using object detection in real-time using neural networks. Particularly, there are 2 of most great importance. The first is an imitation network, which is actually a relatively more new concept, and the second is a vision network, which of course has already been in existence for a very long time, but is now being applied to different fields. So why care about these 2 networks?
Well, the article later goes on to say:
“For example, the blocks didn’t need to be in the exact same location as the demonstration for the system to know how to stack them. If a blue block went on top of a white block in the demonstration, the system replicated that task, even if the starting locations of the blocks wasn’t identical.”
Essentially, the two previous ideas are able to reach an equilibrium together where the system is able to demonstrate stacking. This breakthrough has never been thought of or implemented before, but as you can see, it uses two relatively familiar concepts to deep learning computer scientists and adds them together in a creative way.
To conclude, there is a lot to learn about by seeing some of the more upcoming developments in AI. To learn more, of course, make sure to follow the blog and keep learning! More content is always coming and this is only the first!