Mentalese Anyone?

Posted on Posted in Uncategorized

Funny, even the spell checker does not know what I am talking about. Been reading some books recently in the field of cognitive science. One of them, Benjamin Bergen’s “Louder than Words” kind of got me started with this whole ‘Mentalese’ stuff, but I suspended reading the book for now. I do have some disagreements with the premise in his introduction. I’d love to talk to the author about it. In the book the author was mentioning ‘Mentalese’ as a ‘language of thought’ (LOT), that intrigued me. Upon a bit more research, there is such a concept being discussed in cognitive science circles.
Bergen doesn’t think it has merit, he is pitching a different approach, but in his arguments he has a very limited view on the Mentalese existence and framework. He argues that translating something from one language to another requires both languages to exist. Well nobody knows what Mentalese sounds like, so how can you translate to it. The other thing he advocates is that for Mentalese to exist in the human brain someone must have programmed it.
The narrow aspect here, and my issues with the argumentation is that a human has to learn it’s internal representation over the time of it’s life. So the Mentalese translation is individual. It does not matter if the Mentalese translation varies from one human to another because we all grow up in reasonably similar environments therefore we can understand each other, and if not we can ask.
What I suggest is that the learning process is somewhat programmed, but once a human opens it’s input sensors it can start creating and networking Mentalese concepts and eventually mapping them to a spoken language, or two.
Anyway, I think in building the NeuroCollective I accidentally created a Mentalese recorder. A system where a user can define what he/she means in a word/sentence/idea by linking it to either existing or newly created semantic concepts and then share it with others or compare it to others. I can see a bunch of applications here e.g. where different parties specify what their understanding of a particular subject is and then comparing the Mentalese. So we can, after all, find out is the green I see is the same green you see.
I have by now bootstrapped 3 incarnations of the NeuroCollective, not counting the sandbox test environments. It is literally like teaching a child how to speak, incredibly elaborate and finicky. The bigger idea is to have the data in the system, but the immediate task is to have the code to capture/create Mentalese be as self sufficient and correct as possible. After a bit, the system can be used to assist in sorting out the things it doesn’t know yet or where is sees a conflict with some external data. The more it knows the less supervision is required.

Leave a Reply

Your email address will not be published. Required fields are marked *