Hi, I’m looking for someone with experience in LLM architectures who also has an interest in linguistics or philosophy of language, to collaborate on a research-oriented side-project. The goal is to explore and develop novel LLM architectures capable of genuinely understanding natural language. I have already developed a clearly defined theoretical approach, based on novel ideas in the philosophy of language.
I’m a Java software engineer based in France (I don’t speak French yet, but English is fine). My academic background is in philosophy and mathematics. I’ve developed ideas regarding language and understanding that I would now like to implement. As I don’t have direct experience building LLMs, I’m just not sure how or where to start.
If this sounds interesting to you, please reach out!
Best,
Adam
Welcome @adamrozycki00 Adam. This is your first post. I believe there are many that are interested in Language. Fascinating stuff.
I’d say start reading.
Personally I’m into Gottfried Wilhelm Leibniz and his idea of a universal language we can calculate with. It makes sense for machine life if you ask me.
I started a thread on the subject.
Again Welcome.
Edit: If it pleases you, link I am fairly interested in this, I have been tinkering around with developing an LLM lately this is my repo for attempting to develop a novel reasoning mechanism leveraging multiple mathematical models similar to graph neural networks in terms of math:
https://github.com/Doodle-Med/Mixture-of-geometric-experts Hi, I replied in the wrong thread by mistake, but I’m sure you’ll find my response
Hi @doodle-med , Thanks for the reply! I’ll have a look at your repo, cheers!
Hi again
Thanks for sharing your repo, I read the README. I don’t have experience with the things you do, but from what I could easily gather, you have the skills.
I think that the approach I intend is somewhat different from yours. As I’ve read, you train your model to reason on examples of reasoning. What I intend is to embed the reasoning capability at the architectural level, so that we don’t have to show the model what good reasoning is; it should know it beforehand.
To be frank, I’m specifically interested in a computational implementation of the ideas I’ve already developed about language; not in exploring other ideas or coming up with new ones.
The goal anyways is an architecture that can understand language conceptually in the same way a human does, and besides that, doesn’t hallucinate and can learn incrementally in real time.
I’d like to attach an excerpt of my work so you can get the vibe, but it seems I can’t attach a PDF and I don’t have it published. Basically, I prefer not to share all important details publicly before the job is completed.
If you’re interested, I can share it another way, just let me know. Best
Welcome @doodle-med Thank you for posting and good luck with your project.
Hi @adamrozycki00, I’m working on something similar. Did you make any progress? I’d love to discuss