Video ID: QOYoCwiOyT0
YouTube URL: https://www.youtube.com/watch?v=QOYoCwiOyT0
Added At: 30-09-26 12:01:06
Processed: No
Sentiment: -
Categories: Ai, Technology
Tags: meet, getting, familiar, here, where
Summary
No summary
Transcript
AI agents don't have an intelligence problem. They have a speed problem. LLMs were built to generate text and it works great if you're going to write a document, write code, or just have a conversation. But it is painfully slow when it comes to making decisions. And a software needs to make thousands of decisions at any given point in time. If you're running a payment software, you need to like determine is this a fraudulent transaction. If you're writing a customer support software, you need to determine who do I need to triage this ticket to and you can extrapolate this for any kind of software that you can build. So there's a new company that's come up called Tidesafe. It's started by a researcher named Dieago Almeida who was actually a researcher at OpenAI and built the foundation of GBT and he has now launched a new company called Tidesafe and a new model called Jev. It's uh the etmology of this comes from Jeban's paradox. Now, Jeb is very interesting. Jeb doesn't generate text. What Jeb does is it helps you make decisions. So, let's say you're building a customer support software and you're writing a routing logic. If it's this type of ticket, make sure that it routes to person A. If it's this type of ticket, go and do this autoresponse. If it is another type of ticket, go route this to the engineer. [music] Now, if you are able to put this rubric and give it to Jev, Jev is able to very very quickly route this once it sees a support ticket. Now, why wouldn't you use an LLM for this? Like I mentioned, LLMs are not built for decisions. Jeb is 200 times faster and 400 times cheaper than LLM. So, obviously, it's built for this particular decision-m process. While LLNs were built to generate text, Jev is built to decide where to [music] route from left field. People didn't even expect this kind of an architecture. Follow hackenomics to keep up with the pace of AI.