3 ms·

Towards Grand Unification Theory of AI (Gut-AI)

7 points by kourouklides 4y ago

kourouklides 4y ago
Concurrently with "JEPA" (https://openreview.net/pdf?id=BZ5a1r-kVsf https://openreview.net/pdf?id=BZ5a1r-kVsf), I wrote a very similar (pre-print) paper on *Grand Unification Theory of AI (GUT-AI)*, which is a kind of superset of JEPA, if anyone is interested. I actually made the effort to abstract away complicated mathematics, so that the reader finds it easier to understand, since the quest of AI is a multidisciplinary approach. In my view, I also made better connections to nature (Embedded and Grounded Cognition), among others. I have published it on OSF and since it as a pre-print, I welcome feedback either there or here. Thanks. Paper: https://doi.org/10.31219/osf.io/sjrkh https://doi.org/10.31219/osf.io/sjrkh *PS.* I also made some Github repositories (CC0 1.0 license) expanding the paper and bridging the gap towards practical implementation. - https://github.com/GUT-AI/gut-ai https://github.com/GUT-AI/gut-ai
p1esk 4y ago
Link to a demo?
- kourouklides 4y ago
  Could you please elaborate? What do you mean by a "demo"? An MVP of how it would like? It is actually a work in progress, so it doesn't exist yet. We are gathering feedback on the Research Rroposal, which is in the form of a pre-print paper. Stay tuned for a Whitepaper though, which will be more practical.
  - p1esk 4y ago
    You posted a 126 page “proposal for Grand Theory of AI” on the internet, and want people to read it? If you think you have something valuable to contribute to AI research build a working demo, or at least publish it in a decent conference (such as ICML, ICLR, or NeurIPS). Until then no one is going to take you seriously.
    - gardenfelder 4y ago
      Edit: Violent disagreement
    - kourouklides 4y ago
      Ok, easy tiger. 1) It was intentionally NOT submitted to a "descent" conference (or to a "prestigious" journal that I have already published before as co-author anyway) because I don't believe in that process or system. I used an OSF preprint (which is similar to arXiv, but more properly implemented). 2) Thanks for the feedback on creating a prototype / MVP. What I suggested is related to infrastructure, not so much with the "client-side", so demo is later on in the Roadmap. I just wanted to find some people to join the project early in order to build that demo under the umbrella of a Foundation. I hope it is clearer to you now.
      - p1esk 4y ago
        I’m sorry - I was grumpy when I wrote my previous comment. There are certainly worse ways to spend one’s time than to write about AGI. I have nothing against your proposal (I haven’t read it) and I wish you good luck in building whatever it is you want to build, or finding collaborators.

7 points by kourouklides 4y ago

Concurrently with "JEPA" (https://openreview.net/pdf?id=BZ5a1r-kVsf https://openreview.net/pdf?id=BZ5a1r-kVsf), I wrote a very similar (pre-print) paper on *Grand Unification Theory of AI (GUT-AI)*, which is a kind of superset of JEPA, if anyone is interested. I actually made the effort to abstract away complicated mathematics, so that the reader finds it easier to understand, since the quest of AI is a multidisciplinary approach. In my view, I also made better connections to nature (Embedded and Grounded Cognition), among others. I have published it on OSF and since it as a pre-print, I welcome feedback either there or here. Thanks. Paper: https://doi.org/10.31219/osf.io/sjrkh https://doi.org/10.31219/osf.io/sjrkh *PS.* I also made some Github repositories (CC0 1.0 license) expanding the paper and bridging the gap towards practical implementation. - https://github.com/GUT-AI/gut-ai https://github.com/GUT-AI/gut-ai

p1esk 4y ago

Link to a demo?

kourouklides 4y ago
Could you please elaborate? What do you mean by a "demo"? An MVP of how it would like? It is actually a work in progress, so it doesn't exist yet. We are gathering feedback on the Research Rroposal, which is in the form of a pre-print paper. Stay tuned for a Whitepaper though, which will be more practical.
- p1esk 4y ago
  You posted a 126 page “proposal for Grand Theory of AI” on the internet, and want people to read it? If you think you have something valuable to contribute to AI research build a working demo, or at least publish it in a decent conference (such as ICML, ICLR, or NeurIPS). Until then no one is going to take you seriously.
  - gardenfelder 4y ago
    Edit: Violent disagreement
  - kourouklides 4y ago
    Ok, easy tiger. 1) It was intentionally NOT submitted to a "descent" conference (or to a "prestigious" journal that I have already published before as co-author anyway) because I don't believe in that process or system. I used an OSF preprint (which is similar to arXiv, but more properly implemented). 2) Thanks for the feedback on creating a prototype / MVP. What I suggested is related to infrastructure, not so much with the "client-side", so demo is later on in the Roadmap. I just wanted to find some people to join the project early in order to build that demo under the umbrella of a Foundation. I hope it is clearer to you now.
    - p1esk 4y ago
      I’m sorry - I was grumpy when I wrote my previous comment. There are certainly worse ways to spend one’s time than to write about AGI. I have nothing against your proposal (I haven’t read it) and I wish you good luck in building whatever it is you want to build, or finding collaborators.