For a current-day LLM, deviation could be measured as the number of tokens since the agent touched on the prompted topic. A specific <<refocus>> token could even be trained so that a harness or other tooling could direct the LLM back towards the prompted topic if needed.
Or something roughly like that. Generally speaking the goal would be for the LLM to bring the topic back on its own while incorporating the tangent, but a hardcoded, "Let's get back on track" button would probably be needed
An LLM has a fixed number of ways it can express itself. we can give it an array of 14 billion options but it still has to chose one to output. Humans have no such limitation.
An LLM does not persist in consciousness from one token to the next. Each generation, happening hundreds of times a second, will be initialized, generate an output, and terminate. Humans are not stateless like an LLM.
Raines Law in that late 1800s put restrictions on the sale of alcohol in NY on Sundays. Bartenders used a loophole of the drinks being served with a meal to get around it.
They would make a single sandwich, "serve" it to the patron along with their drink, then immediately take back the sandwich to "serve" it to the next person wanting a drink
You're browsing for whatever piques your interest, and the library wants to curate the collection based on what people are interested in. The books that collect interest get placed on the shelves and the ones that don't get archived. If it's in the archive it probably wouldn't have interested you.
>They didn't force this person through the scanner
"Despite the woman's request for a pat-down search, a TSA agent told her that her only option was to pass through the AIT device."
>they could have asked for a supervisor and discussed the situation
"Before passing through the device, the plaintiff spoke to another officer, trying to explain the situation, but was told that the AIT machine had been “adjusted” so that it would not damage her spinal cord implant."
I’m in agreement with the blog post. I’ve been treating AI more like a tool and less like a science experiment and I’ve gotten some good results when working on my various side projects. In the past much of my time was taken up by research and learning the various little parts of how everything works. What starts as a little python project to play around with APIs ends with me spending 5 hours learning tkinter and barely making any API calls.
A little late, but this is something that I've been considering a lot lately. When there's a limited resource (funding) how do you determine who will receive it?
For something like this I think a citizens assembly[1] may work best. Take all artists receiving funding and are NOT up for renewal. Select a number of them randomly to form the assembly. This assembly then reviews submissions from artists up for renewal and determines if they meet a minimum standard for funding to be renewed.
Or something roughly like that. Generally speaking the goal would be for the LLM to bring the topic back on its own while incorporating the tangent, but a hardcoded, "Let's get back on track" button would probably be needed