Hi Korrelan, Nice project. Impressively implemented.
But I'm interested in exploring assumptions in it. Like you I want to get below levels of abstraction (no cars from cars..?) So I'm attracted to that aspect of your project. But if we reject historical abstractions, we need replacements. I'm interested to know exactly what bases of abstraction you suggest as a alternative. Something emerging from your model of the connectome, I guess. That sounds like your guiding principle: only assume the connectome, don't assume any abstraction of it?? But even trying to base everything on the connectome, are there any (car like?) assumptions in the abstractions you end up with? As a test case you cite your replacement for the phoneme: On Wed, Jun 26, 2019 at 4:13 PM korrelan <korrelan...@gmail.com> wrote: > I require a frequency modulated speech/ phoneme engine, none exist so I’ve > just started building my own. > > https://www.youtube.com/watch?v=ffY37q44O4E > The video you link says the CMU-Sphinx phoneme dataset. So in this demo you are using existing abstractions. But you say you want to get away from them: "The KorrTecx model learns speech by listening to its own vocal output and then compares it to what it has heard from it's peers, but I can botch phonemes for now." > Which sounds like you want to replace phonemes with some kind of protocol based on "what it has heard from its peers". What's that? How would it differ from current phoneme models? It sounds like you consider phoneme models to be artificial (cars?), and want to replace them with something more from first principles? What would be the basis for this structural principle to replace phonemes? How will your system break up "what it has heard from its peers" and produce something to replace the "botch" of phonemes (to be coded with frequency?) Is that structural principle repeated structure? That's my suspicion. Will you simply be looking to identify patterns which repeat in "what it has heard from its peers", and use those repeated elements as a first principles replacement for the "botch" of phonemes? -Rob ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf97c751029c2e4db-Mff193925c8fe2319e2ff9933 Delivery options: https://agi.topicbox.com/groups/agi/subscription
