Live data from Hacker News

What it feels like to work in AI right now

robotic.substack.com

521–530 of 768 posts

Re: What it feels like to work in AI right now

#521

From my perspective it’s just _confusing_ to work in AI right now. We have some massive models that are doing some really neat stuff, and apparently hundreds of millions of people are using them—but I keep wondering: to do _what_, exactly? I’m not asking what the models can do, I’m asking what people want the models to do every day, all the time. I’ve been shown some neat pictures people made that they thought were c…

Most of the people asking the question "to do _what_, exactly" seem to have not much of a "knowledge worker" experience. Every non-tech person I showed GPT4 to was after a few tests immediately using it to speed up parts of their daily work. Most impactful is how it destroys the blank page/getting started barrier, second how easy it is to substantially change/adapt/refocus produced work. It is like having an incredib…

> Most of the people asking the question "to do _what_, exactly" seem to have not much of a "knowledge worker" experience.

From personal experience I'd say the opposite: GPT lacks the specialist knowledge to produce useful writing or yield accurate answers in any of the markets I've worked in (I'll grant that less niche markets exist, and that GPT is pretty good at fixing the writing of people that lack English language writing skill) and it seems like people egging GPT as replacing most of those roles are all showcasing hypothetical "generate a website for an imaginary product with minimal brief" kind of situations which GPT excels at because there aren't any real world knowledge worker constraints imposed on their solution. That's definitely not to say it has no use, but lots of less technologically impressive accomplishments like data entry wizards and templates also have use without being considered transformative.

Re: What it feels like to work in AI right now

#522

Earlier quoted context omitted.

Amen. It outrages me that someone would waste my time making me listen to a regurgitated, averaged-out, speech written without any emotional fire or depth at an occasion that is profoundly meaningful. A speech as a best man should be personal, heartfelt, it should be funny and a little cringy maybe, and it should mean something. Same with eulogies; this is a moment to celebrate someone's life, to state what that pers…

That's quite arrogant. Not everyone has ever written something like it. Why is it unfair or unpersonal if chatgpt helps you, guides you and gives you a good starting point? Guess how many people would Google some example and starting with that, what's so wrong to let chatgpt generate a shitty first draft already tailored to your situation?

Hello greeny7373, nice to meet you.

> Why is it unfair or unpersonal if chatgpt helps you, guides you and gives you a good starting point?

As somewhat extreme examples: Rodin didn't swoop in and smooth off "The Kiss" after an apprentice chiselled the basic outline out of a lump of marble; pretty sure Salvador Dali started with blank canvas instead of getting a basic landscape from a $2 Art Shop and added a melting clock and a giraffe to it.

We all have language.

Take your age, subtract maybe 5 years, and that's how much experience you have expressing yourself in your language.

Say you're 20; by that metric you have 15 years experience communicating. Now, I don't play guitar and I'm not remotely musical, but I'm pretty sure if I did it every day for 15 years I could at least bash out somthing original, if possibly influenced by things I liked.

I'm not suggesting that people entertain us for half an hour with heartfelt, witty speech about their relationship with the groom; I'm not suggesting ten minute poetic ode to a life well-lived that leaves everyone simultaneous trying not to cry and trying not to laugh, and nodding and saying "yes, that's how they were".

It doesn't need to be long: just one memory or incident. It doesn't need to be Shakespeare: just heartfelt. It doesn't need to win the Academy for Best Actor. It doesn't need to be a 5 paragraph essay (though if you did use that model, there's nothing wrong with it).

It just needs to be you; it just needs to be yours.

(I already know what I'll say at my Father's funeral when the time comes: it'll be about two sentences, just something he said to me once. But I know that's all I need to say).

Maybe I am arrogant, but I want human feeling and human expression at human events. I don't want to be snoring through regurgitated pap.

Re: What it feels like to work in AI right now

#523

I think what’s really depressing here is just how effective scaling seems to be. It just means that any company that’s not willing to pour hundreds of millions of dollars into their AI programs isn’t serious at all and would probably be better off hiring engineers to figure out how to integrate GPTX into their systems than trying to roll their own. I really think we’re going to see a massive collapse of AI/data scien…

I wonder if there’s an incentive for a large group of companies to fund open source models, sort of like Linux.

Re: What it feels like to work in AI right now

#524

Earlier quoted context omitted.

Hmm…no, I’m not doubting the value of the interface. I’m asking what’s the “killer app”. Maybe I should just ask ChatGPT to explain it to me…

It’s programming. The thing is really great at programming, and everything else is icing on the cake. They already got my $20 a month lol

Not really programming but generelly producing text (or maybe generally certain structure like ASTs at some point). Programming is a subset.

I expect that this will move lots of things into the world of text that weren't before.

Re: What it feels like to work in AI right now

#525

Earlier quoted context omitted.

I mean Google revolutionized search. Apple revolutionized personal computing. OpenAI revolutionized… rewriting things with slightly different wording? I’ve seen so many breathless people posting “this would have taken me so long to search” and then I type 3 keywords from their massive prompt they crafted and find it instantly on Google. We’re talking 1000x or more faster. I feel like the same is happening in your com…

You can program GPT in English. Let me repeat that: You can program GPT in English . ENGLISH! You're complaining about the first nuclear test bomb being impractical and uninteresting. How will this change the world? That huge monstrosity had to be affixed to the top of a test gantry and took years of effort by a veritable army of the best and brightest to make! No way it could change war, or geopolitics, or anything.…

> Let me repeat that: You can program GPT in English. ENGLISH!

For problems that are fine being defined ambiguously. Try to program a database in English, let's see where it goes.

Re: What it feels like to work in AI right now

#526
In the past, we had notable individuals in various scientific or mathematical fields who had multiple successive breakthroughs throughout the duration of their career (Einstein, Feynman, Gödel, Hilbert, etc.)

With ML/AI, that doesn’t so much seem to be the case. There are the early pioneers (Hinton, LeCun, Bengio, etc.), but it seems more as though they were the first to “discover” neural networks (that actually worked), and then the individual breakthroughs sort of stopped after that. This observation is not a jab at these people—rather, it’s because I wonder if in machine learning, unlike the more traditional areas of math and science, one person is just not able to test groundbreaking new ideas on their own anymore. A lot of the latest progress in ML comes from large companies consisting of teams of researchers who are largely unknown to most of the public.

I’m not quite sure what my point is, but personally it’s a bit sad to me that fundamental development in AI now appears to require a vast amount of resources that small teams or individuals don’t have access to. I suppose you could argue this is a similar situation to Bell Labs, but even in that case there were many distinct contributions from well-known individuals working there.

Re: What it feels like to work in AI right now

#527

From my perspective it’s just _confusing_ to work in AI right now. We have some massive models that are doing some really neat stuff, and apparently hundreds of millions of people are using them—but I keep wondering: to do _what_, exactly? I’m not asking what the models can do, I’m asking what people want the models to do every day, all the time. I’ve been shown some neat pictures people made that they thought were c…

It's just good enough and approachable. But I agree, for what I do professionally it just doesn't even come close and I'm already quite efficient.

From me and around me:

- marketing asked me to show them the ropes of mid-journey yesterday, boss said "this will be the face of our new product" to one of my hasty "creations".

- mom learns english with chatGTP because she finished duolingo

- I wrote a PoC demo for a prompt engineering tool and gpt4 demo chat in a day. That would have taken me days without gpt4. (material design, storing in local storage, gobbling up the data specification, save everything on change not with buttons)

- First draft for some diagrams in mermaid js worked well too, or converting from flow diagram to a swimlane diagram.

- All kind of personal data cleaning: dirty list of emails > ready to paste in mail client, wall of text with broken new line characters > sub-headlines and paragraphs

- virtual assistants (e.g. company chat-/voicebots), needs a bit of tooling but gpt4 is totally ready for it (apart from latency and price)

- 30 minutes to a browser-add-on that marks tweets as "seen" so I can skip them if I scrolled past them before. (userscript to be precise)

- understanding tax regulations %)

What I'm waiting for:

- better knowledge ingestion so it can use my notes

- personalization over time

- good dev-ops integration (push and deploy for me too).

- maybe something to have better separation of concerns on code so the messiness matters less, not sure if possible :)

- remote control my screen

- running an LLM locally

- have it build its own plugins for any website or service

Re: What it feels like to work in AI right now

#528

AI world can be proper dystopian these days. I know someone who accepted a job offer in one of the biggest market research firms in the world. She was making AI models for a high tech company before and was hired to make summarization AIs. Between serving out the (frankly ridiculous) 2 month notice period and joining this new job GPT-4 was launched and prices came down for 3.5. The original plan for building somethin…

Could be ideal for next season of Bored to Death.

Re: What it feels like to work in AI right now

#529

Earlier quoted context omitted.

That's just not true. I expect software to become a magnitude shittier soon because of this attitude. After that I expect software to become unbelievably good, because thanks to AI we will have the ability to prove correctness of software much more cheaply than before, and to design on a higher level than before. After all, you don't want the AI to help you generate boilerplate code, you want the AI to help you avoid…

It is true if you can understand the code it writes. Maybe you are worried junior programmers are gonna churn out code using ChatGPT but to be honest I'd rather trust code from ChatGPT than from a junior, I feel ChatGPT writes better code on average.

I'd argue you only THINK you understand the code. If it generates a 1000 lines, which look like they are doing the right thing, will you be diligent enough to go through every single one of them? This literally can only work for extremely boilerplate code (which is maybe the code most of programmers write), but for most code I write I need a mental model of it, and constructing that model myself is easier than to try to learn it from some code. Of course ChatGPT can work as an inspiration, especially for working with unfamiliar APIs.

Re: What it feels like to work in AI right now

#530

Earlier quoted context omitted.

I feel that there’s significant insecurity among a lot of coders about GPT-4. A lot of them are ignoring the pace of improvement and highlighting the few off chances where it gets things wrong.

I've not used GPT-4, so it could be different, but regular old GPT-3.5 gets a _lot_ of things wrong.

GPT-4 is leaps ahead, and it's improving with every new release. The latest March 23 release is significantly better than the previous one and does a LOT of heavy lifting for my code at least.

At the very least, it's a massive productivity booster.

Post reply on HN