The speaker is discussing a project to solve a 5,000 piece all-white puzzle using a robot. The robot is designed to take each piece one at a time, take a picture of it with a high-resolution camera, and then extract the shape of each piece from the photos. The speaker then plans to use these shapes to determine which pieces could potentially fit together. This involves dividing the perimeter of each piece into four sides and then comparing these sides to see which ones might fit together.
The speaker also mentions a technique called "locality sensitive hashing," which is used to find all the similar edges for all the pieces. This technique involves creating a 2D or 128-dimensional graph and using it to calculate the sequence that describes where each point falls on the graph. This allows the robot to directly look up what other points are similar without having to do any comparisons.
The speaker also talks about a vibratory Bowl tumbler, which is used to smooth the edges of the puzzle pieces. However, this process has resulted in some pieces changing shape and causing issues with the puzzle's assembly. The speaker has to rescan these pieces and recalculate their shapes.
The speaker also discusses how to handle pieces that don't fit together perfectly, which sometimes happens due to a "Cascade" effect. This involves moving pieces around until they fit together correctly.
Finally, the speaker discusses the robot's ability to solve the puzzle. When put into solve mode, the robot expects to be loaded with pieces one by one. It extracts all the details of each piece, looks up where it needs to go using the solution, rotates the part to the necessary orientation, and then moves over and puts it down where it needs to go.
The speaker concludes by saying that he hopes the video has inspired people to become programmers, as this is a fascinating job where one can explain to a computer in detail how to do things like this.
1. The speaker is attempting to solve a 5,000 piece all-white puzzle, which they believe would take an average person a very long time, potentially years.
2. The speaker has a robot that can assemble puzzles, but it's not very sophisticated and can only handle around 50 pieces.
3. The speaker has been working on an algorithm that will hopefully allow the robot to handle larger puzzles, specifically those with 5,000 pieces.
4. The speaker is frustrated with the complexity of all-white puzzles, as there's no picture to help figure out if pieces will fit together.
5. The speaker has calculated that to figure out which pieces connect, they would need to do around 30 million comparisons.
6. The speaker has made hardware upgrades to their robot, making it better, faster, and stronger.
7. The speaker has considered moving their workshop and has faced challenges related to this process, such as moving a delicate puzzle-solving robot.
8. The speaker has a computer made by a company called Puget systems, which they use for their work.
9. The speaker has been working on finding a way to directly compute which edges match a given edge without having to do any comparisons.
10. The speaker has used the technique of locality sensitive hashing to optimize their algorithm.
11. The speaker has discovered that their puzzle pieces have been changed by a vibratory bowl tumbler, which has altered their shape.
12. The speaker is considering making their own puzzle to avoid further complications.
13. The speaker is planning to feed the pieces into the robot one by one, as they have not yet built an automatic feeder.
14. The speaker has encountered an issue where sometimes when pieces are put down, they don't fit together perfectly.
15. The speaker has managed to solve the puzzle using their robot, despite some challenges along the way.