What it feels like to work in AI right now • Hacker News Board

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

#491

Post by girvo » Fri, Apr 07, 2023, 9:26 AM UTC

kolinko wrote: post #417

nwienert wrote: post #204

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…

As for googling, here are some examples of queries you can try and see how it works: - summary of all the carbon neutral concrete methods, especially ones that can be done in a small industrial workshop as a prototype - I have allergies in Thailand, mid-february. What may it be related to? - list all the companies from Japanese stock exchange that have high debt rate Those are top of my head, but really anything that…

How do you deal with it straight up lying? My problem with this whole system is, if I’m asking those questions it’s because I don’t understand the field well enough to answer it myself, which means I can’t pick up on if ChatGPT is lying…

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

#492

Post by TheMode » Fri, Apr 07, 2023, 9:28 AM UTC

maxilevi wrote: post #236

TheMode wrote: post #179

Earlier quoted context omitted.

I do not believe that standard of living should be compared in a vacuum. Since I assume that this is what you are referring to, I am firmly convinced that happiness is relative. What this means is that improving the standard of living of the average person by 5% while improving it by 50% for a smaller population will hurt more than it helps.

Why would it hurt?

Ignoring study, that would be a more graphic example: https://www.youtube.com/watch?v=meiU6TxysCg

Or historically, if this wasn't true we would be the happiest humans to ever live on earth. Doesn't seem to be true on any possible metric.

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

#493

Post by nwienert » Fri, Apr 07, 2023, 9:29 AM UTC

eitland wrote: post #335

nwienert wrote: post #204

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…

> and then I type 3 keywords from their massive prompt they crafted and find it instantly on Google. Seems I and you have different Googles and you still have the one I had pre 2010. For over a decade now, Google has been including things I never asked about to the point where it would sometimes be easier to find it using Marginalia. Some say it is just because internet has changed and there is less ham and more spam…

I actually agree Google has gone downhill. Yet for the 8 or so examples I’ve tested where I saw hyped GPT results, every single one google answers, usually in the top snippet, always in the first result.

For politics shopping and some other topics it can be terrible, but I don’t think GPT is good at those either.

I’m actually happy to be proven wrong here. If you have some examples let’s test it out. If it’s a true step function improvement I’d expect it to be easy to source examples.

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

#494

Post by rntz » Fri, Apr 07, 2023, 9:33 AM UTC

> All of these low-level concerns make working in AI feel like the candle that burns bright and short. I'm oscillating between the most motivated I've ever been and some of the closest to burnt-out I've ever felt. This whiplash effect is very exhausting.

My candle burns at both ends;

It will not last the night;

But ah, my foes, and oh, my friends—

It gives a lovely light!

(First Fig, Edna St Vincent Millay)

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

#495

Post by weatherlite » Fri, Apr 07, 2023, 9:35 AM UTC

JohnFen wrote: post #53

davesque wrote: post #28

I feel like the whole ChatGPT bubble has really highlighted what feel like some fundamental shortcomings in the worldview that is represented in tech. That is, there seems to be a winner take all dynamic baked into the tech world. Maybe this arises from the simple fact that tech places a lot of power in the hands of individuals. But there's an emergent downside to this which is that it makes those who were already mu…

> there seems to be a winner take all dynamic baked into the tech world. This has been a pathology in the computer/software industry for a very long time. It's never been actually true except in a couple of special cases, but the industry acts as if it is. That has led to all sorts of bizarre and undesirable things. > Guys, what the heck are we doing?? I think we're playing with fire and, without extreme caution and…

What's your main concern with all of this?

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

#496

Post by rlt » Fri, Apr 07, 2023, 9:36 AM UTC

version_five wrote: post #455

rlt wrote: post #437

Earlier quoted context omitted.

> Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith. https://news.ycombinator.com/newsguidelines.html > questioning the value of a chatbot Original commenter was questioning the value of the entire field. > hasn’t had a real job” This is different than “haven't worked in a 'real' business”.

Ironic given yours and others emphasis on my choosing "chatbot" to summarise llms as a reason to dismiss my comment, along with the rest of the pedantry. The upstream post dismissed / insulted the person for questioning value, which was what I called out. If you really had wanted to get into the "HN rules" game, you could at least have cited "don't be snarky"

You weakened the argument by portraying it as more insulting against a narrower claim than it was.

But yes, also snarky.

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

#497

Post by notahacker » Fri, Apr 07, 2023, 9:36 AM UTC

SanderNL wrote: post #337

onion2k wrote: post #324

Earlier quoted context omitted.

That's very cool, and right now it's a good idea, but I strongly suspect GPT only looks clever and does a good job in isolation . If everyone starts using it product announcements will start looking very similar, and they'll lose a lot of their impact. This is definitely the case with cover letters for jib applications. The ones written by GPT appear to be pretty obvious - my guesses could be wrong, but after seeing…

I find it interesting it’s very hard to make it do typos. The type of errors humans make. It can do typos but even those seem weird. If I push it to make errors it either over- or undershoots. The “correctness” of it is a definite give away IMO.

In fairness, lack of typos in a cover letter has traditionally been interpreted as a sign of diligence!

For more general queries its "house style" tends to be really obvious, with all it's "however, it's important to note" and "ultimately it depends on"s and the tendency to flesh out a one sentence answer to the specific question with two paragraphs or five bullet points of detail at a slight tangent to it...

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

#498

Post by zone411 » Fri, Apr 07, 2023, 9:38 AM UTC

Analog24 wrote: post #266

zone411 wrote: post #154

Earlier quoted context omitted.

I'm quite surprised at how little progress FAANG companies have made in recent years, as I believe much of what's happening now with ChatGPT was predictable. Here's a slide from a deck for a product I was developing in 2017: https://twitter.com/LechMazur/status/1644093407357202434/pho... . Its main function was to serve as a writing assistant.

Scaling up an LM from 2017 would not achieve what GPT-4 does. It's nowhere near that simple. Of course companies saw the potential of natural language interfaces, there has been billions spent on it over the years and a lot of progress was made prior to ChatGPT coming along.

You're making incorrect assumptions. This project wasn't about scaling any published approaches. It was original neural net research that produced excellent results with a new architecture without self-attention, using a new optimizer, new regularization and augmentation ideas, sparsity, but with some NLP feature engineering, etc. Scaling it up to GPT-2 size matched its performance for English (my project was English-only and it was bidirectional unlike GPT so not a perfect comparison), and very likely scaling it up to GPT-3 size would have matched it as well, since GPT-3 wasn't much of an improvement over GPT-2 besides scale. Unclear for GPT-4 since there is very little known about it. Of course, in the meantime, most of these ideas are no longer SOTA and there has been a ton of progress in GPU hardware and frameworks like PyTorch/TF.

You can check out my melodies project from a year ago as a current example. There is nothing matching it yet: https://www.youtube.com/playlist?list=PLoCzMRqh5SkFPG0-RIAR8.... And that's just my personal project.

What you're saying about companies recognizing the commercial potential is clearly wrong. It's six years later and Siri, Alexa, and Google Home are still nearly as dumb as they were back then. Microsoft is only now working on adding a writing assistant to Word, and that's thanks to OpenAI. Why do you think Google had to have "code red" if they saw the potential? Low-budget startups are also very slow - they should've had their products out when the GPT-3 API was published, not now.

One thing I didn't expect is how well this same approach would work for code. I haven't even tried to do it.

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

#499

Post by IanCal » Fri, Apr 07, 2023, 9:39 AM UTC

cgearhart wrote: post #152

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…

Scrolling back through some discussions with chatgpt and a few things with bing (later ones more likely to be gpt4)

* Explaining what a do/while(false) loop was for as I'd not seen that construct before

* Discussing what DIDs were and how web DIDs worked as the RFC was very detailed in a lot of areas I just didn't care about. A discussion with a pretty knowledgeable person in the area was what I needed, and what I got. It explained what the well-known part was, explained my confusion around resolving to a document and resolving to a resource (where I was mentally stuck).

* Creating a learning plan, diagrammed by mermaid, for progressing in bouldering. Each major step broken down into sub-parts to practice

* Finding https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem given bad explanations that didn't lead me to the right place in google

* Finding https://blockprotocol.org/ given a shaky memory of "some composable ui framework by someone famous" iterating a little with bing. I had failed to find it before manually.

* Explained and created a table of bouldering gradings as the place I go uses a different one to the videos I see

* Discussed project ideas to do with my son, gave me great ideas around electronics that I think are a good fit as well as a few other things. The most useful part here was being able to say "that's too generic, I need proper projects" and "that's too simple for him" and have it update. It then also created some good explanations at different levels about how radios work,

* General discussions about long term impacts of LLMs, potential use cases

* Career advice

* Generating art that we'll be getting commissioned for the house

* NER without any coding

* Generating ember templates, CSS and example filling data for a custom framework given a problem statement (what I'm actually building right now)

* (edit) I just took requests from my kids and made them some colouring in pages with robots, firefighting robots, lego ninjas, owls, frogs and crabs.

> In fact, no one in my extended sphere of friends and family has asked me anything about chatGPT, midjourney, or any of these other models. The only people I hear about these models from are other tech people.

I've had family members tell me they've used it to create reports, and used it to create marketing copy, a website and lecture slides for others.

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

#500

Post by girvo » Fri, Apr 07, 2023, 9:40 AM UTC

Fr0styMatt88 wrote: post #407

cgearhart wrote: post #152

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…

I’ll give you a concrete example from a few days ago in my job. I needed a quick utility window in the Unity editor to see what animations could fire animation events and what those were. I’m somewhat familiar with the editor API, enough to know what to Google and roughly where to go in the docs. I don’t do it enough though to really learn it beyond that point. So I’d estimate I could spend maybe one and a half hours…

One worry I have for the long term is: how will we learn or adopt new programming languages in the future, when LLM du-jour knows nothing about Language X? Will we be stuck with what we’ve got because LLMs make us too productive in them?