Inspiration
I'd like to talk about my story. Well let's start with how and what I inspired from: Since my 15, I was interested in philosophy, and once when I came across Liars Paradox, I started to solve one of the fundamental problems of philosophy. But hours, days, months passed, no idea, no results. Suddenly one day, I came up with division the information into 3 sets: True, False and Undefined, after which they great idea came to my mind. This is contextual operator K, which was created to measure the amount of context and in which the information intersects. That's how I started to develop my system.
What it does
Since this filter-website should be integrated with AI, a person gets in ChatGpt (example), writes a request, and filter analyses the data in it, and searchs whether the information matches with the global sources or its database. If it does, it just confirms it, because the information in the filter is sent to True set. If it doesn't, it debunks or rejects to agree with that, since the statement is delivered to False set. But if it is unknown, AI tells the truth without any hallucinations and probability tokens.
How we built it
As I mentioned before, the system has a Contextual operator. This operator is also divided into 3 types, K+, K-, K_0 each of which operates along with the set their sign matches. So how it works: if K=K+, then S (Statement)€T, if K=K-, then S€F, therefore if K=K_0, then S€U. Moreover each contextual operator has its boundary: K+€+Z, K-€-Z, K_0=0. One thing which makes my system dynamic is the formula "K(t_1)+Delta_K=K(t_2) which shows how a certain information changes over time
Challenges we ran into
What's interesting here is that the only problem I encountered was that the limitation of the AI tool I'm using which is Claude. Claude in its base mode is very restricted by low limitations, because of which it took more a lot of time than it was supposed to be.
Accomplishments that we're proud of
However, despite this fact, I'm truly fascinated by the fact that I managed to create such useful technology.
What we learned
Therefore, from all the experience I have gone through, I learned how to create websites using different AI tools, which, I think, will be beneficial in my professional life
What's next for Anti-hallucination filter
Actually, one of my greatest goals is to integrate my filter with big AI companies and future AGI. So, if I will succeed, the next thing I want to focus on will be my another projects that I left on the foundation level
Built With
- claude-api
- cssanthropic
- html
- javascript
- python
- wikidata-api
- wikipedia-api
