Live data from Hacker News

First Impressions with GPT-4V(ision)

blog.roboflow.com

181–190 of 345 posts

Re: First Impressions with GPT-4V(ision)

#181
post #60

Earlier quoted context omitted.

Has AI been able to figure out conflicting and quickly changing product requirements and implement them on top of existing technical debt? You'll be fine.

People keep saying "ah but it can't do X!". So what? Most of us have multiple decades until we can retire. This AI is getting better every few months. It will be able to do it even faster, better and more cheaply than a human can.

> It will be able to do it even faster, better and more cheaply than a human can.

Take what you did in the past year. Write down every product decision taken, every interaction with other teams figuring out APIs you had, all the infra where your code is running and how it was setup and changed, all the design iterations and changes that had to be implemented (especially if you have external partners demanding it).

Yes. All that you'd have to input into the AI, and hope it outputs something decent given all that. And yes, you'll have to feed all that into AI all the time because it has no knowledge or memory of "on Monday the new company bet was announced in the all hands"

You will be fine.

Re: First Impressions with GPT-4V(ision)

#182
post #2

I’m intrigued to see what kind of problems it’s going to be good/bad at. I think it’s going to be tricky to evaluate though because it has probably memorized all the easy images to eval it with. Eg anything pulled from Google Images (like that Pulp Fiction frame or city skyline photo) is not a good test. It recognizes common shots but if you pull a screenshot from Google Maps or a random screen cap from the movie it…

It seems it will still be limited by its linguistic understanding of the surrounding context, at least in the first chicken sandwich picture. Although its interpretation could make some sense but is also mostly wrong if talking about physical size of a modern GPU's main processor compared to the size of the associated VRAM chips. It has missed the joke entirely as far as I am aware. I think the joke is actual about N…

Ironically the exact processors need to run GPT-4V in the first place.....

Re: First Impressions with GPT-4V(ision)

#183

Earlier quoted context omitted.

That's a new flavor of customer service jobs which will still suck and be paid terribly

At least you'll be able to complain to someone, hehe.

Unless that person is you... but in that case you would just complain to your coworkers so nothing really changes I guess

Re: First Impressions with GPT-4V(ision)

#184

It didn't successfully explain the NVIDIA burger joke though? The image is making fun of how nvidia has implemetned price discrimination by releasing consumer gpu's that don't have as much vram as they should so that they can sell the fully loaded datacenter gpu's at insane markup without pissing off gamers. The explanation didn't even come close to getting that.

Yeah that was my take as well. Although it did come up with a plausible sounding answer. A less nerdy human might also not get it

To be fair the author didn't get it either so hard for me to blame GPT but was a bit wild to see the first example of success actually be a failure.

Re: First Impressions with GPT-4V(ision)

#185
Back when they first announced this months ago they leaked the demo Discord bot ID in some logs they showed on screen, which you could use to invite it to your own server. For a while you could use it through that until they shut it down, though it was incredibly slow and went down a lot

Re: First Impressions with GPT-4V(ision)

#186

Earlier quoted context omitted.

> they will become a superior user interface to almost every thing you want to do No they won't. They're actually a pretty terrible user interface from a design perspective. Primarily because they provide zero affordances, but also because of speed. UX is about providing an intuitive understanding of available capabilities at a glance, and allowing you to do things with a single tap that then reflect the new state ba…

If something is a repetitive habit that you can do almost without thinking, there is a good chance an AI could infer that entire chain. I think what's more likely is that an AI based interface will end up being superior after it has had a chance to observe your personal preferences and approach on a conventional UI. So both will still be needed, with an AI helping at the low end and high end of experience and the mid…

Like on Apple Watch Ultra, when I press the start exercise button, the exercises that it suggests are in an order unique to my baseline activities.

Re: First Impressions with GPT-4V(ision)

#187
post #9

I'd be interested in knowing how good it is at solving visual captchas, do we foresee a huge rise in automated bypasses?

We have just added a section on this! TL;DR: GPT-4V isn't great at this task at the moment :)

Back when they leaked it via a Discord bot I found it worked better when you ask it to first describe each box

Without doing that: https://cdn.discordapp.com/attachments/964175221089259591/11...

With it: https://cdn.discordapp.com/attachments/964175221089259591/11...

(though it's only one example so it could be coincidence)

Re: First Impressions with GPT-4V(ision)

#188

It didn't successfully explain the NVIDIA burger joke though? The image is making fun of how nvidia has implemetned price discrimination by releasing consumer gpu's that don't have as much vram as they should so that they can sell the fully loaded datacenter gpu's at insane markup without pissing off gamers. The explanation didn't even come close to getting that.

It explains the broad joke, but gets the labels wrong.

Says that the small buns are labeled "GPU and VRAM" and the massive fried chicken is labeled "NVIDIA BURGER"

Should say that the small bun is labeled "VRAM" and the massive fried chicken is labeled "GPU"

Re: First Impressions with GPT-4V(ision)

#189
post #10

Sure, there are a few edge-case failures and mistakes here and there, but I can't help but be in awe . AWE. Let me state the obvious, in case anyone here isn't clear about the implications: If the rate of improvement of these AI models continues at the current pace, they will become a superior user interface to almost every thing you want to do on your mobile phone, your tablet, your desktop computer, your car, your…

So you won't be able to do anything without Internet connection to the AI mainframe? No thanks.

At some point Open-source models' performance will follow. Llava will improve.

Re: First Impressions with GPT-4V(ision)

#190
post #43

I’m impressed, technically, but this seems niche. Who holds their phone up and takes a photo then wants to know it was a photo of? That’s weird. If you don’t know what it is, wtf did you take photo? The obvious use here is natural language improvement / photo editing for photos, but this is just a stepping stone to that, and bluntly, as it stands… the examples really don’t shine… Great for the vision impaired. …not s…

> who takes a photo then wants to know it was a photo of? I find myself doing this rather frequently. The scenario described in the article is quite common for me: capturing a photo of a plant and utilizing an existing classification service to determine its identity. It could be driven by mere curiosity or practical concerns like identifying whether a plant is poison ivy. Wildlife identification also falls into this…

> Wildlife identification

Wouldn't say this is super reliable, I gave it a photo of a small squid in my hand and it said it was a baby fish (very obviously was not a fish).

Post reply on HN