Jev – the mini revolution in the AI world
AI as you know it has some serious problems: hallucinations, indecisiveness, and it's designed to generate text. That's nice, and at the same time often useless for automation. And then there was Jev.
Jev is precisely not an LLM but an AI model that can make structured decisions at lightning speed — exactly what LLMs can’t do, and exactly what LLMs need. Jev should perhaps be seen more as a companion AI that quickly determines the correct, pre-defined result. Think for example of support tickets that are automatically routed to the right place. So not: “Yeah no, Tina from accounts handles that” — instead, the ticket comes in, gets sorted, and goes straight to Tina.
So if you have a million support requests per day (think for example of Microsoft, where things go wrong fairly often), those requests can be routed to the right place many times better and more securely. And that’s a meaningful automation — something LLMs can’t do, and certainly not at that speed.
In fact, a book has been written about this: Daniel Kahneman, “Thinking Fast and Slow.” Here, “thinking fast” describes the intuitive, rapid thinking and acting on it — think of opening a door — which Kahneman calls System One. System Two describes the slow, thoroughly considered and multi-angle thinking, which is what LLMs typically do.
Jev is therefore a complementary model and absolutely cannot “chat.”
The creators of Jev — of whom there are now several claiming to be the inventor of this form of AI model — say themselves that Jev could, for example, sort thousands of emails in record time. An LLM would read the emails, think about what’s meant, who it’s intended for, and then where it needs to go. In short: that’s not exactly efficient.
Jev is the intuitive, pre-programmed thinker that can.
Another example: we live in the age of AI, and you notice it on your firewall too. Attacks are increasingly automated and AI is being used more and more to carry them out. That means you ultimately need AI to defend yourself. And that’s possible. Over the past few weeks, we recorded a baseline on the basis of which we let an AI decide which attacker should be blocked. Pretty cool, if we say so ourselves. This is also a perfect example of what Jev can be used for: lightning-fast determination and decision.
Sensor data. Another area we frequently encounter. For reading and sorting these kinds of small 4K data blocks, Jev would also be a valuable tool. And content moderation — which the techbros say is truly impossible — could be greatly improved with Jev.
We wouldn’t be surprised if the major model makers quickly integrate this into their models and we move towards a “Society of Mind” where different parts are deployed in a targeted manner.