OpenAI's Gabriel Petersson argues future jobs will focus on instructing AI, claiming programmers already spend 80% of their time doing so
Aaron Levie argues smarter models cannot replace human context.
Commenters backed the Box CEO's focus on context and proprietary data to make AI work well plus simple ways to share goals, yet called naive the idea that explaining tasks to AI will take most work time.
Most Activity
There’s no amount of intelligence that can get packed into AI models that replaces the need for context. For any sufficiently general purpose AI, you will always have to guide it in the direction you want as it has an infinite range of directions it can go in.
As long as the same model is used by a lawyer, an engineer, a financial analyst, or a healthcare professional, and as long as you’re trying to do anything uniquely differentiated or specific, then instructions, domain context, and proprietary data will always need to get into the context window for the model to be useful.
This is partly why AI automation doesn’t come for free, and why there’s still a wide spectrum of who’s getting the largest gains from AI and who’s not. You have to put in real work, and you get real value on the other end.
This is one of the advantages that applied AI will also have in the market. Any layer of abstraction above just the raw intelligence that can meaningfully get you off to the races faster will likely continue to be valuable.
every job will turn into explaining your intentions to ai
explaining what you want to ai is surpringly time consuming, coders already spend 80% of their time doing it, and this will be true for everyone
@powerbottomdad1 try delegating a product feature to someone who has never worked on a product before in less than 5434 sentences
maybe real agi can make decisions, but for the next 8 months we're in a state where the models can do all work but cant make any decisions
if we had gpt 9 and it was 100x more intelligent it would NOT be a simple textbox. it would be a hyper sophisticated app that let's you communicate with it with much higher bandwidth than text in text out
we thought intentions can be explained with a sentence
but there is so much more context to every intention
tell the smartest person in the world but who doesn't know you to "clean your inbox", they'll get so many things wrong
and it would take you hours of walking through emails to write a full 10 page instruction for what exactly is relevant and what is not, that updates daily
@powerbottomdad1 or try tell a dumb (or smart) person to "clean your inbox", they'll 100% guaranteed remove items you wanted and miss removing things you don't wanted. no intelligence gets around this
context and personalized models will at some point, but we'll do most work with ai before then
@gabriel1 Taps sign*
@gabriel1 real agi wouldn't need that much explaining given that very dumb people can infer a lot from very vague instructions
@powerbottomdad1 yeahhh i had your exact opinion before, took a long time until just recently until i realized that the trend of interfaces is actually going towards being significantly more specialized the smarter the models become instead of the other way around
@gabriel1 Super Intendo
yes x1000 been saying this a lot - its literally a communication skill issue. reminds me a lot of like a hybrid pm/engineer function. those that can communicate their needs best with the model, state intentions provide necessary context etc etc will be the most valuable humans on earth
@gabriel1 i'm talking about telling a construction worker to "install the window" or "weld the pipe". not saying they're dumb, just that they don't need lots of hand holding. i do think that smart people can clean an inbox pretty well tbf. but agree for a while its about writing the spec
@gabriel1 i mean yeah i agree
@gabriel1 @powerbottomdad1 In grade school our teacher made us write instructions for "How to make a peanut butter and jelly sandwhich"
She would then take your instructions and do them word for word. Thus, you realized your instructions were inferences.
That was a good lesson in how to prompt LLMs.
@gabriel1 This take seems very weak. A true AGI should understand your context, intentions, and needs better than you can explain them yourself
@gabriel1 Prompting becomes the new profession.
@gabriel1 on the upside this is basically what coding is anyway
@gabriel1 AI whisperer, the best whisperers will be those that have have a joint psychosis with AI and come out the other side with wisdom, consciousness permanently altered. Cognition, secure, but ready to traverse the latent space at FTL♾️
@gabriel1 To me GPT is part of a subsystem that other intelligent systems work with in symbiosis. We have so many systems in nature that communicate and don’t involve a focused high bandwidth brain. Zoom out. Shape the environment for intelligence to thrive.
@gabriel1 Whiteboard in 3D AR while also Neuralink chip and camera shit. I really need ai to see my full screen + webcam seing me and talking to me and visualy showing things on another screen and Im not even starting here. We need 6D connection to this latent space