Hacker News new | past | comments | ask | show | jobs | submit | nkingsy's comments login

The example of the cat and detective hat shows that even with the latest update, it isn't "editing" the image. The generated cat is younger, with bigger, brighter eyes, more "perfect" ears.

I found that when editing images of myself, the result looked weird, like a funky version of me. For the cat, it looks "more attractive" I guess, but for humans (and I'd imagine for a cat looking at the edited cat with a keen eye for cat faces), the features often don't work together when changed slightly.


Chatgpt 4o's advanced image generation seems to have a low-resolution autoregressive part that generates tokens directly, and an image upscaling decoding step that turns the (perhaps 100 px wide) token-image into the actual 1024 px wide final result. The former step is able to almost nail things perfectly, but the latter step will always change things slightly. That's why it is so good at, say, generating large text but still struggles with fine text, and will always introduce subtle variations when you ask it to edit an existing image.

Has anyone tried putting in a model that selects the editing region prior to the process? Training data would probably be hard, but maybe existing image recognition tech that draws rectangles would be a start.

Genuine question - how would such a model "edit" the image, besides manipulating the binary? I.e. changing pixel values programmatically

I think the unstated assumption is that there's a block list somewhere being fed into a guardian model's context


The chain is spazzing out in every shot like it's tightening and loosening over and over again. Mid drives already wear out chains quickly. Seems like this is just asking to snap a chain


There is lots of human feedback. This isn’t a game with an end state that it can easily play against itself. It needs problems with known solutions, or realistic simulations. This is why people wonder if our own universe is a simulation for training an asi.


It would be out of date in months.

Things that didn’t work 6 months ago do now. Things that don’t work now, who knows…


There are still some tropes from the GPT-3 days that are fundamental to the construction of LLMs that affect how they can be used and will not change unless they no longer are trained to optimize for next-token-prediction (e.g. hallucinations and the need for prompt engineering)


Do you mean performance that was missing in the past is now routinely achieved?

Or do you actually mean that the same routines and data that didn't work before suddenly work?


B

Each new model opens up new possibilities for my work. In a year it's gone from sort of useful but I'd rather write a script, to "gets me 90% of the way there with zero shots and 95% with few-shot"


The word swinger wasn't mentioned once. Probably because the swingers are just quietly enjoying their lives under the radar.


I think it might also be because "Swinging" is a word from a previous era and some/many of the young LGBTQ+ people are against learning from their elders.

Swinging is a very clear example of ENM.


> young LGBTQ+ people are against learning from their elders

No, there's a missing generation. As (previously) one of those, my generation is now the "elders", and we had to learn in a strange, weirdly sheltered way. Our elders were dead or hiding. The topics were taboo, the representation garbage, and the content online? Often blocked in the place we had internet access.

I do a lot of teaching to my younger queer friends. Sometimes I have to do a lot of research on a topic before I can give an answer.

In addition, Swingers weren't talked about in any part of my growing up. It wasn't until as an adult I looked at my partner and said "Oh, they're having a key party" at an exhibit of a model 50s-era home in the midwest.


Sorry, yes, I wasn't intending to give the impression that I was ignoring the many, many that died in the 80s during the AIDS epidemic. I was intending to refer to Millenials as the elders and Gen Alpha as the "young LGBTQ+".

At least what I see on TikTok has reflected a lot of the elder (Millenial) LGBTQ+ people becoming periodically frustrated with the younglings for not listening to them as they talk about exactly the frustrations / issues the new ones are going through or aren't having to go through.

Weirdly, swingers *were* talked about in my family when I was young.


Google seems to think monogamish is the new word for it, but that is a really confusing word (I thought it meant you can cuddle puddle with your friends).


ENM?


Ethical non-monogamy


> some/many of the young LGBTQ+ people are against learning from their elders.

Well almost a complete generational cohort of their elders is simply missing. They died of aids in the 80s and 90s.


No, they didn't. At least look at the data before making such a bigoted statement.


Yes, we did lose a generational cohort.

There is a reason that the quilt is so large[1].

[1] https://www.npr.org/2012/06/27/155868611/pieces-of-aids-quil...


100k out of 200 million (US)? Out of 5 billion (worldwide)?

The flu makes us lose five "generational cohorts" every year, then.


Intersectionality.

100k out of *how many out homosexual individuals during that era in the area being studied*?

And it's not just raw numbers, either. It's how many lives / families were impacted in this unique way.

To add, it's also 100k that were almost entirely in a single demographic that was explicitly and implicitly being harmed by those in power during that era.


I want to add more. The whole “30 is gay death” expression, I bet, especially since it’s not really a thing anymore, was greatly influenced by there being very few gay people over 30. Because the rest were dead.


>Intersectionality.

Nah, just elementary school Math.

It was around 1%, a considerable morbidity but still very far from losing "a complete generational cohort".


Hey btw I absolutely did mean "a generational cohort of lgbtq people" not the broader total population. Which would make no sense, so I assumed no one would interpret my meaning that way. Thanks for calling me a bigot and then starting a fight with these other people about it though.


A multiplication and a division is starting a fight nowadays?

>a generational cohort of lgbtq people

That's where the 1% figure comes from. If you look at the whole population it is more like 0.0001%.

Sorry if Math hurt your feelings.

Btw, bigot means a person who is obstinately or unreasonably attached to a belief or an opinion, which seems relevant.


His little bike helmet is adorable


The AI safety team was really proud of that one.


Western POWs. Eastern front was a whole other thing.


Fair enough!

(And if you want a good history of horror in broadly-speaking early-middle 20th century Eastern Europe, I suggest reading Bloodlands by Timothy Snyder.)


Me? I downloaded a wurly patch “wurly logic” that I can pretty much play like a piano on the reface dx.

I bring it camping all the time.


Hard to say as someone with the power.

Ask a bird what flying is good for and their answer will be encumbered by reality.

Kind of the opposite of “everything looks like a nail”.


Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: