Thinking Machines
This article proposes a truly comprehensible AI architecture: a deterministic thinking machine based on concrete concepts. The author suggests building a recursive descent parser for the entire English language, where each word triggers a function, combined with a reasoning engine and response generator, to achieve precise understanding and computation of language.
Notifications You must be signed in to change notification settings
Fork 0
Star 0
Copy path
More file actions
More file actions
Latest commit
History
History
History
22 lines (12 loc) · 1.62 KB
Copy path
Raw
Copy raw file
Download raw file
Manual-AI
The first truly comprehensible AI
AI has taught us one thing - we can understand and know the universe. it turns knowledge and reason are not magic, but rather concrete entities that can be computed.
All this talk about AI has changed the way we all work and think. We no longer view our brains as these godlike objects, but are now able to have humility and see that we have become obsolete.
But we have yet one more frontier to cross - a true thinking machine. A thinking machine operates on distinct concepts, not abstract probabilities.
Instead of modern AI which is solely statistical, a true thinking machine is deterministic. I propose the following architecture as human's last original thought before we are truly obsolete.
I propose a true recursive descent parser for the entire english language. Every word will have its own function to parse and disambiguate homynyms. All language has definite meaning therefore there are a finite number of rules, but too many for one person to write.
Once we make this parser, we need a reasoning engine that goes through the tree and knows exactly what in being asked and exactly what needs to be computed.
The idea is all language has a finite set of meanings, therefore proper responses are not magic, but rather programmable.
The idea is each word triggers a function, for example where triggers a function that searches temporal memory and knows the rules for finding data.
Finally the response generator takes the computed answer and logically determines the most correct output syntax.
With AI this could be done fast but what if we all did it manually as our last hurrah!