4 ms·
The best analogy I can think of is that if you want your agent to accomplish something without persisting a long chat history, but instead use the agent to reor
by WiSaGaN 3y ago
The best analogy I can think of is that if you want your agent to accomplish something without persisting a long chat history, but instead use the agent to reorganize and change the prompt, you would choose to use the method Leonard uses in the film "Memento". Due to his condition, Leonard cannot form new memories and struggles to recall events that occur after his injury. Leonard knows his condition and uses tattoos to record facts "between sessions". Each chat completion of the LLM is similar to "each session" Leonard experiences. The prompt, with the help of the agent, can persist across LLM chat completions, which is similar to the tattoos, and various notes and photos he has.
- Hugsun 3y agoExcellent analogy
- nerdponx 3y agoYeah but there's a plot twist wherein it's revealed that the method has some severe flaws, and it's not as effective as Leonard thinks it is.
- Widdershin 3y agoNever considered that Memento was a cinematic adaption of Reflections on Trusting Trust.
- tbrockman 3y agoWhich so satisfyingly parallels prompt injection attacks.
- ilteris 3y agocontext size is still going to be an issue when history gets large, correct?
- abecedarius 3y agoSee also https://en.wikipedia.org/wiki/Soldier_of_the_Mist https://en.wikipedia.org/wiki/Soldier_of_the_Mist
- ianand 3y agoAndrew Mayne (ex-OpenAI prompt engineer) actually uses Memento as a great metaphor for describing LLMs https://x.com/ianand/status/1723976526436417546?s=46&t=x6aIMKxtOHuyj50nx-e74A https://x.com/ianand/status/1723976526436417546?s=46&t=x6aIM...