Acumos Project's 1st Software, Athena, Helps Ease AI Deployment
IBM has an AI platform also, as does MS, Google, and the rest. At the heart of all these things is a kind of 'join' system with various grammars for finding paths through the spaghetti. I should work with these folks, Acumos is Open Source. But I fear bloat, I fear the case where everyone has their own analysis tool jammed into the system making it difficult to sort the good and bad.
My best bet is to plug away slowly and post the files. Soon I will have competitors trying to architect the same idea, incremental grammars operating on structured data.
I will win, I have no bloat. The core join program has less than a thousand lines of code, before grammars are loaded. It is not a GUI framework, it is a machine framework, designed to be crammed into a AI co-processor. Developers are free to add the GUI that supports their needs, but a python application built over the join machine is the goal.
the join system is built around LazyJ, which is an executable form of data and execution, in one format based on comma separated file, like JSON. By adding grammar specific operators to join, the LazyJ syntax is still usable, it is context adaptable and works on high and low lever data structures, like templates. LazyJ is relatively undefined in terms of its search,replace and delete capability inside the join. The added functionality I add to LazyJ is all experimental, ad hoc. I fully expect the pros to get a complete LazyJ definition soon. Ayway, I have the winning architecture.
I will win, I have no bloat. The core join program has less than a thousand lines of code, before grammars are loaded. It is not a GUI framework, it is a machine framework, designed to be crammed into a AI co-processor. Developers are free to add the GUI that supports their needs, but a python application built over the join machine is the goal.
the join system is built around LazyJ, which is an executable form of data and execution, in one format based on comma separated file, like JSON. By adding grammar specific operators to join, the LazyJ syntax is still usable, it is context adaptable and works on high and low lever data structures, like templates. LazyJ is relatively undefined in terms of its search,replace and delete capability inside the join. The added functionality I add to LazyJ is all experimental, ad hoc. I fully expect the pros to get a complete LazyJ definition soon. Ayway, I have the winning architecture.
No comments:
Post a Comment