Samuel 🦋: yeah I naïvely thought that nobody would care about using a vision model, since image classifiers seem non controversial to me? I wouldn’t dare touch a diffusion model since I agree those suck in every conceivable way. anyway clearly people don’t see it that way

Deleted.

yeah I naïvely thought that nobody would care about using a vision model, since image classifiers seem non controversial to me? I wouldn’t dare touch a diffusion model since I agree those suck in every conceivable way. anyway clearly people don’t see it that way

“AI” has become a catch-all term for the entire ML field and imo different parts are different amounts of bad

Yeah, I personally don’t trust OpenAI, given their cagey comms on how they source training data. Even if they claim they won’t train with data sent to their API. But definitely agree that “AI” has taken on a genericide-like usage, akin to calling all gaming systems “a Nintendo.”

I also feel bad for the “generative art” community, which has to endlessly explain that they use math and deterministic algorithms to generate patterns and it’s not at all related to genAI.

Every term other than "large learning model" tries to pass it off as some other unrelated thing that already exists to give itself legitimacy, which has the side effect, intentionally or not, of making those other things sound illegitimate. A tactic for products that are totally real and not scams.

Yes, and the first plane to strike proc gen art was "NFT" shit! It's unfair that they now have to distance themselves from a new fraud that has nothing to do with them. bsky.app/profile/symb...

Something similar happened to generative/procedural art (when NFTs brainslugged everything adjacent to it). Now you have to parse everyone involved to see if they're involved or pushing that garbage.

I actually do agree but I think it’s more than just “AI” that rankled people here, it’s gpt specifically. the former is a triggering word now for sure, but for the latter, I think many people just top-to-bottom don’t like that company

I totally should’ve just used Claude instead, you’re totally right

the most ethical thing to do here is obviously to start your own nuclear reactor using a bluesky microgrant

bluesky chip fab when

should have used a local phi3 mini

sam's laptop has been through enough as is

Shoulda used a Geiger counter

At the very least, you should have told people BEFOREHAND

but that would never label the great unwashed!

Well, now I'm worried. What part is worse than Sam Altman's planet killing scam factory? Is there an actual SkyNet happening out there somewhere?

AI-driven fighters that can outfly human pilots.

yeah honestly sucks, I'm really interested in ML myself, but now I'll have to be super careful with wording to not immediately trigger everyone who just knows the umbrella buzzword

It's a shame the picture and text generation is called AI. It has completely muddled the topic making reasonable discussion about AI impossible.

…how did people think the classification would work if it didn’t use ML lol

I guess that’s a silly question, there’s definitely a lot of required knowledge for understanding the solution spaces for these problems and not everyone can be expected to have that knowledge

I don’t think people realized it was an image classifier. I naively assumed (never assume!) it was a random number generator or maybe a deterministic hash of a display name or something. Samuel updated the description though, so people can make an informed decision if they prefer not to opt in.

yeah I apologise that it wasn't there from the start, lesson learnt

Hey, you had a great idea, people clearly love the Kiki vs Bouba (faux?) rivalry. OpenAI and GenAI in general are just very polarizing topics and full-stop deal breakers for many people. 🫣

that I certainly have discovered!

I’m sorry you had to discover that the hard way 😅 I hope your week goes better from here!

I'm no programmer or anything, but I did program an dice roller in Q basic in high school. Wouldn't it be way less time, bandwidth, and resource intensive to just have a number generator?

For sure! RNG is much less resource intensive. If the goal is to get the same answer for the same user every time, another low-resource option would be something like a hashing algorithm. But idk what the goal was in the implementation, if some analysis of an image was the whole point, etc 🤔

luckily pinging something like GPT4o (which is designed to be really efficient relative to prev models) consumes very few resources to the point where it's negligible, and libraries out there make it very easy for the developer to build. doing the actual classification part prob took 1hr or less

afaict, the process is: 1. send a very low res version of your avatar along with a message that's like "based off these definitions of kiki and bouba, , what do you think this image is more like?" 2. API responds with "looks kiki" or "looks bouba" 3. labeler applies label

Hmm, I’d be curious how you’re defining “negligible.” Surely the CPU, GPU, and memory consumption of a GPT4o generation is significantly higher than generating a random number or hash (some hashes will vary). GPT4o may be fast but resource consumption is still very high.

It’s not a traditional image classifier (like an MLP network or whatever), though, it’s still using a large generative multimodal model. There’s nothing wrong with that, ofc, but the model architectures and techniques used are extremely different.

yeah, just from people talking about it I assumed it would be a consistent hash of the DID with a pepper

Yeah same here, never even thought it would be an image classifier. I thought it was completely random and that you could re-roll if you unliked and liked again.

Image classifiers are definitely dangerous in a lot of applications, including medical and obviously any deployment of "facial recognition". There are countless examples of both, but here's just one to start with. venturebeat.com/business/whe...

When AI flags the ruler, not the tumor — and other arguments for abolishing the black box (VB Live)venturebeat.com One of the big issues that exists within AI generally, but is particularly acute in health care settings, is the issue of transparency.

Are you concerned that people are gonna waste time at the doctor because they were incorrectly diagnosed with Bouba

the risks are pretty clear from that example.

god, it's weird to be so anti-ai and be the voice of dissent here but ... this is quite literally just a fun toy, not afaik a propaganda piece for OpenAI

not at all. Ref: "image classifiers seem non controversial to me" to be clear, image classifiers ARE dangerous, and I gave some examples.

What is the danger in using an image classifier for a silly social network labeler?

it's a fun thing and honestly I appreciate you letting folks know that you used OpenAI + that OpenAI _shouldn't_ be ingesting this data. it sucks OpenAI has poisoned the terms ML/AI so much, because there are genuinely amazing use cases for these tech stacks that aren't "poach artist stuff"

Of all the truly bizarre and incoherent defenses being offered, this might be the most incoherent: "Yeah, it sucks that OpenAI has been poison for these terms but it's cool of you to use OpenAI for a fun toy that harms no one."

I mean I'm not really defending the use of OpenAI here - I thought the concept was fun and image classification is a fun way to do it. I just mostly appreciate the OG author letting folks know and expressing that I hate OpenAI ruining AI/ML in general. Not trying to pick, just being honest /shrug

Bold of you to assume most people complaining know the difference...

Bold of you to pretend the difference is meaningful in any way that matters Get mad at OpenAI for sucking up all the oxygen in the room, and get at Sammy BlueSky for giving them more oxygen. He should apologize, not pull this naive babe-in-the-wood routine about why people are angry at him.

All you smarmy fucks are providing rhetorical cover to the very people who have tarnished the image of the machine learning field. Maybe irreparably, I don't know. But it's not going to be repaired by you all having a sensible chuckle at the peasants who don't want the planet burned down for scams.

What's wrong with diffusion models tho? I love me some stabdiff horrors in the morning

Post