Reliable email automation starts with legible decisions

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 16:00

   And then scenario two, I'm sending this to a GEV-powered system. What happens? Compare and contrast. Can I do an easier example? Then customer service is actually quite complicated. Let me do something that is very similar to customer service, but is a lot easier to think about. Like imagine an executive assistant for your email. It's kind of like customer service, but the amount of things you can do is way simpler.

2. 16:26

   And as far as I can tell, I could be wrong. that like a scheduling assistant is still not solved. I, you know, I have emailed a lot of people. I don't see automated scheduling systems. And like, that's kind of weird because it really looks like LLMs should solve this, right? So people have been trying for years for this and how you do it in an LLM is you would, you know, make a big system message write-up of all of your preferences, blah, blah, blah.

3. 16:51

   And they would be incredibly good at stylistically being accurate. like it'll never not look like a well-written thing but it will often make subtle mistakes not in the style not in the sycophancy or sycophancy i don't know how to pronounce that word um you know like it'll always be people-pleasing because lms are so good at people being people-pleasing but it may not make the right decisions and what people will do is they will like keep programming their system prompt like please do this if it's an investor do that If it's inbound email, do this, you know, blah, blah, blah.

4. 17:26

   So what ends up happening is you're basically programming in natural language, this giant system prompt. And as the complexity rises, the error rates get higher and higher and higher because you're just praying model, please don't make those mistakes. And what tends to happen is it's embarrassing when those mistakes are made or it's costly. You know, you may not have like scheduled something with an important investor or something like that.

5. 17:47

   and then you need to like rip the system out and toss it away and just use you know a human because you know it's like it's just obvious what to do there and they can like ask you when they're uncertain or whatever else so llms to my understanding still have not solved this problem though i don't know if anyone has solved this with a jev like system yet but i can compare and contrast the differences of it so what you can imagine instead of programming in natural language like describing like if it's important blah blah blah if it's from this blah blah blah if it's inbound blah blah blah um you can actually make individual classifiers for each of these things like smart if statements like you can now express things that you couldn't express in software before because how can like these bit comparisons figure out if it's inbound or not right like

6. 18:36

   that's an intention and it requires like natural language common understanding and the idea is you can make classifiers for all of these things and you have this legible system where not only can you measure if every if statement is accurate you know that none of those if statements will be forgotten and if ever a mistake is made you can add it to an eval because the system is very legible so it'll probably start out less good than the text-based system because the text-based system has a lot of sane defaults you know it looks very reasonable because that's what it's trained to do it's this train to like like look human pleasing but it actually doesn't have any of the actual behavior you truly care about and you don't have the control over that you know and that is a major problem for anyone who wants to truly have something run in the background

7. 19:24

   they want it to not worry about it and they want to have control and they want to have like some idea that it's not going to just go completely off the rails and that is what combining like a you You know, system one models with code will enable you to do because code is kind of like, you know, the roller coaster tracks. You don't go off the rails. Code prevents that from happening. And similar to that type safety, you know, that is what the property that the system one models have.

8. 19:53

   And these things also prevent going off the rails. And because they're very legible, it becomes way easier to build these. Just trust that they can run in the background.