[drakmo]

Google is just a lightning fast but messed up index combined with really good graphs. ChatGPT now is only one tool Soon it will be just one piece in a box filled with a combination of tools that let you just know whatever you want to in a depth of detail you specify it to be laid out:

- large conversational language model that generates the piece of information

- fact checking model that applies post checks on logic and consistency

- recitational graph model that backs the propositions with sources and ratings

- geospatial and locational model that adds references in space

- recognitions development model that can classify how realizations and knowldege develop over time

- social trend model that can adjust to conversations of the society as a whole

All this in combination will bring any knowledge in any level of detail with a defined level of certainty.

Do you want to know more?

[janalsncm]

I would like to know more about the fast fact checking model. This will require some kind of symbolic NLP approach that doesn’t exist yet.

[drakmo]

Here is a good recent overview of where automated fact checking is positioned right now: https://aclanthology.org/2022.tacl-1.11.pdf Fact checking of AI generated content can potentially be optimized for this use case. And for speed: I think many of the computations for several layers of AI processing can be decentralized in the future and re-establish the heart of the Internet. The contributions for such computations need a high level of trust and therefore a public immutable ledger might be a viable approach. So perhaps decentralized AI might be the use case where finally some "blockchain-crypto-cringe" might actually be useful ;-)

We need to be aware that we just entered the steep part of the exponential growth graph. Amazing times ahead.