I've been inspired by Light Sail 2, and wrote some science fiction (possible).
Some believe deep learning will help the symbolic AI with its messy data problem, creating hybrid systems that will achieve AGI. What they fail to recognize is the focused (and interpolative) nature of deep learning algorithms make this impractical. The problem with AGI is the breadth of knowledge it requires. A hybrid AGI would require at least as many power-hungry deep learning models as exist today. Combining the failings of the symbolist and deep learning approaches can only lead to more failure. That's not to say symbolic investigations have been completely fruitless; they've lead to the development of object oriented programming which is useful for creating AI.
It's time for a fresh approach to AI, one based on ideas rather than the size of your processors. This is the second in a series.
There are plenty of tools, many of them free, enabling programmers who want to create an intelligent robots; all that's needed is a PC and some skill. I'm posting the first of two short papers discussing a road to intelligence (the one less traveled).
Robots, and other types of anthropomorphic beings, have been playing out roles in mythology since medieval times, but it turns out robots may need mythology, just as we do, to communicate.
Category theory can be applied in different fields in mathematics and extend functional programming languages such as haskell, but can also be used to describe a vocabulary used by a conversational robot. Previously, I used category theory to describe a vocabulary's fundamental building blocks. I'll continue by using category theory to describe how a vocabulary can be developed into a category a conversational robot's software can use.
I'm becoming interested in applying category theory in development of my conversational robot. I've attached a short document describing categories in vocabulary. Category theory is a new discovery for me.
Machine learning and other AI algorithms have difficulty recognizing relationships between people, and other concepts we humans take for granted. Ri is software I've developed that uses semiotics to address these issues. If you show a picture to a machine it may be able to recognize shapes in the picture, but it cannot tell what's going on. Walid set a challenge illustrating this problem (https://lnkd.in/gncF3xt). He provides two pictures: one in a classroom and the other in a home setting. The challenge is this: A 4-year old asked to point to the picture with the teacher and the picture with the mother will correctly identify the pictures. Provide software that can do the same. Ri talks with users and can embellish a conversation with context from stories in its library. For this challenge I use the "ri reply" command to remove that user interaction variable and leave everything up to the software. You can find the results in the attached doc. I've also provided a comparison with Google search.