Hard AI

My current strategy is two-legged.  The first leg is to find papers with benchmarks that current LLMs struggle with, and show my Turing complete LLM can solve these. The second is to show that it can solve challenges that current LLMs are good at. One of these things is that LLMs are really good at answering questions if they have enough knowledge of a problem. If we ask ChatGPT how it have knowledge about cities it answers this: " I rely on knowledge encoded from geographic and linguistic data up to my training cutoff (mid-2024), which includes global place name databases such as Geonames , Wikidata , national census gazetteers , and academic or development sources (e.g., UN and World Bank geographic datasets). So when you mentioned “Boulma,” I recognized it as matching entries from Burkina Faso , where several small villages with that name appear in those official datasets. I didn’t search the web — it’s based on general geographic knowledge and structured data I was traine...

The previous experiments have shown that a path some of the critical components for ASI are possible is https://hardai-omnia.blogspot.com/2025/08/asi-experiments.html. But this page tries to better explain the overarching framework, since many pieces are still missing in the implementation.  I am open to ideas what to train next. To be precise what it is we are describing. And LLM, that works exactly like a normal LLM. Except it is extremely good at retaining details and use those from a very context window. And when training it will optimize both for correctness and performance. But it is limited just like a normal LLM, and should be put into an agentic framework. So agentic frameworks that does RAG, coding, connect to databases and so on, are still very much needed. But this LLM will use whatever you put in the context window in the best way possible.  We should probably also touch a little bit about how AI companies will make money in the future. There is a race to the bott...