TechTakes
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
Prompt fondlers come full circle and re-invent Eliza:
I suffer from monkey mind, chronic imposter syndrome, and the hedonic treadmill. A few friends have confided that they have the same issue too. Perhaps you can relate as well. To improve my mental well-being, I meditate, write morning pages, and keep a gratitude journal. However, as helpful as these techniques are, they primarily rely on introspection. What was missing was an external perspective—someone who could ask follow-up questions, challenge my assumptions, and help untangle my thoughts and emotions.
please, for the love of fuck, see a real therapist for the ADHD and depression (and the hedonic treadmill? which just seems to be your mood regulating back to normal after bad shit happens?) you’re desperately trying to cloak behind pseudopsychological terminology
Enter ChatGPT. Over the past year, I’ve been using it as my therapist and coach. It’s been surprisingly helpful and meets all the needs above, plus it’s available 24/7. And I don’t have to worry about being judged (hopefully our AI overlords will be benevolent!) So far, every “session” has been insightful and I almost always leave with a lighter heart.
there’s so much to unpack in “I don’t have to worry about being judged, haha hopefully the AI gods won’t hate me haha” alone
System Prompt: You are Tara, an empathetic, insightful, and supportive coach who helps people deal with challenges and celebrate achievements.
You have academic and industry expertise to brainstorm product ideas, draft engineering designs, and suggest scientific solutions.
but what good is a therapist if it can’t help you hustle and grind! certainly stealing ideas from a magic 8 ball will help you with your chronic imposter syndrome!
fucking hell I just clicked to their about page
Hi, I'm Eugene Yan. I design, build, and operate machine learning systems that serve customers at scale. I also write and speak about ML, RecSys, LLMs, and engineering.
I'm currently a Senior Applied Scientist at Amazon where I focus on helping customers read more. Here, I built systems including real-time retrieval, bandit-based ranking, and recsys in search (see RecSys 2022 keynote). More recently, I'm exploring how LLMs can help us serve customers better.
Outside of work, I share the ghost knowledge of applying ML via ...
I guess there's the minor upside that the people making the world worse are all so proudly self-identifying? removes the need to guess
Yeah, it'll make it easy to identify them for re-education after the revolution...
System Prompt: You are Tara, an empathetic, insightful, and supportive coach who helps people deal with challenges and celebrate achievements.
You have academic and industry expertise to brainstorm product ideas, draft engineering designs, and suggest scientific solutions.
but what good is a therapist if it can’t help you hustle and grind! certainly stealing ideas from a magic 8 ball will help you with your chronic imposter syndrome!
you know, that shit reads exactly like that the affirmations scenes in severance, and that shit was made to be as creepy as it was
it's an extremely sad testament to a myriad set of societal failures that this person is finding depth in the text of a soulless sample-glitching regurgitation cloner that's using inspiro-copy from however many self-help websites as the pattern for this. like, this is some fucking suckerpunch-level suck for shit to be this bad for that person
MOM: "We have Torment Nexus at home"
https://www.theregister.com/2024/04/03/stability_ai_bills/
"Stability AI reportedly ran out of cash to pay its bills for rented cloudy GPUs Generative AI darling was on track to pay $99M on compute to generate just $11M in revenues"
"That's on top of the $54 million in wages and operating expenses required to keep the AI upstart afloat."
Midjourney is profitable, so they fucked up BIG time. Imagine going bankrupt while someone else uses your technology to make money hand over fist.
lol it's real https://twitter.com/Grimezsz/status/1420822651442921472
I think there's a level of US cultural stuff I don't understand here but I'll just sneer away, hoping no-one quizzes me on it.
I recall reading some of this around the time, seeing the rest of it is even more wild. Talk about a bad trip, heh
I now regret asking for and getting this information.
how long have you been here
the frequency of terrible TILs here is higher than anywhere else
Promptfondler proudly messes with oss project (OpenAI subreddit)
To be clear nothing in the post makes me think they actually did what they are claiming, from doing non-specific 'fixes' to never explicitly saying what the project is other that that it is 'major' and 'used by many' to the explicit '{next product iteration} is gonna be so incredible you guys' tone of the post, it's just the thought of random LLM enthusiasts deciding en masse to play programmer on existing oss projects that makes my hairs stand on end.
Here they are explaining their process
It's code reading and copy pasta.
Amazing that when there is pushback against his ideas he starts to ad hominem call people professors.
(there was this talk/article about how to recognize a crank I forgot the link/source but iirc 'reacting badly to pushback' and 'always trying to solve the biggest open problems firsts' are two things on the list).
The halting problem relates because in a similar way it is saying that there cannot be a single algorithm that solves everything within the system thus the system needs algorithms that can do "work" to solve complex problems. NP-complete problems are out of reach for today's classical compute systems and thus quantum computing could approach them and unlock them faster i.e. the speedup. This does reflect Godel's incompleteness theorems.
I need a shorter name for the whole genre of person that’s on way too many uppers and won’t stop using ChatGPT all night to make all their decisions, cause we keep running into them and for some reason all of them are obsessed with CS woo
maybe we’re really just witnessing what happens when your “nootropic” habit gets out of hand and you’re still in debt from gambling on crypto, so you get high as shit and convince yourself you’re a genius because whenever you read the tea leaves they tell you exactly what you expected to hear. this is, unfortunately, how cults tend to form.
“Presumably-outsourced non-thinker” but unfortunately it’s not as snappy as promptfans and promptfondlers
Prompt-deciduous?
They are having a delirium supported by ML, so Promptlerius?
At least they get called on being a word salad merchant in that sub.
So, all this tells me is that GPT5 is going to be scary good and I can't wait.
Amazing how much tech hype nowadays is 'the next version will be great!'. Parts of this always have existed, and there is also the other part of tech hype 'You don't get it, this isn't just tech, this allows you to be A Platform!'. Vast fields of new possibilities, always just out of reach. Fusion is 17.6 years away people!
E: Related to that, also see how people always need to shift to the next big thing. The next codebase will fix your problems, no the next new AI system will be better, dump the old and learn the new thing. (Don't forget to not notice you are not actually doing things, just learning new systems over and over).
What do you mean “fixed” an entire repo? How were you prompting and what were you fixing?
crickets
there are errors and you read them into the ai. Someone alerted that a build wasn't working.
Somebody set up us the bomb. Main screen turn on.
When you're refactoring you need to be more familiar with the code base. For example, why are you refactoring you ask yourself. What parts of the code do that functionality. How is it intertwined to other parts of the code base.
Figuring out a codebase from first principles.
That person is getting paid twice as much as you now for using chatgpt to code. And they still don't know what #include means.
I don't even get what the guy is claiming to do here. Hope he didn't try sending his fixes upstream.