Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OpenAI's new open-source model is basically Phi-5

seangoedecke.com

Post reply on HN
Print view View original

131–140 of 233 posts

Re: OpenAI's new open-source model is basically Phi-5

#131

Post by cess11 »

numpad0 wrote: post #128
wmf wrote: post #20

I saw a bunch of people complaining on Twitter about how GPT-OSS can't be customized or has no soul and I noticed that none of them said what they were trying to accomplish. "The main use-case for fine-tuning small language models is for erotic role-play, and there’s a serious demand." Ah.

Maybe I'm just not seeing it, but is that use case really real and not just prudish hallucinations? The market for NSFW novels is smaller than even cyberpunk paperbacks, there's no way everyday people build up addiction for an interactive version of it.

The claim that Fifty Shades of Grey and, I don't know, Norman Mailer?, Chuck Tingle?, are less common on book shelves and night stands than cyberpunk paperbacks seems obviously wrong to me.

Perhaps you could elucidate further on this subject? I'm mostly into books from 1800-1985 or so and don't know much about contemporary literary fashion.

Edit: Jean M Auel was extremely common in occidental households a few decades ago, especially the first and second books about Ayla, I'd wager much, much more common than cyberpunk.

Same goes for books by Alex Comfort.

Re: OpenAI's new open-source model is basically Phi-5

#132

Post by throwuxiytayq »

nickpsecurity wrote: post #114
izabera wrote: post #26

Earlier quoted context omitted.

what's the problem with that? we have erotic texts dating back thousands of years, basically as old as the act of writing itself https://en.wikipedia.org/wiki/Istanbul_2461

[flagged]

i think god did a fairly hack job overall and i’ll gladly en masse commit acts that please me and fail to please his non-existent ass.

i’d even turn gay but that’s a bit out of my comfort zone

Re: OpenAI's new open-source model is basically Phi-5

#133

Post by amanaplanacanal »

numpad0 wrote: post #128
wmf wrote: post #20

I saw a bunch of people complaining on Twitter about how GPT-OSS can't be customized or has no soul and I noticed that none of them said what they were trying to accomplish. "The main use-case for fine-tuning small language models is for erotic role-play, and there’s a serious demand." Ah.

Maybe I'm just not seeing it, but is that use case really real and not just prudish hallucinations? The market for NSFW novels is smaller than even cyberpunk paperbacks, there's no way everyday people build up addiction for an interactive version of it.

I suspect NSFW novels is a bigger market than you think. Spicy romantasy is popular among a certain set of readers.

Re: OpenAI's new open-source model is basically Phi-5

#135

Post by mejutoco »

NitpickLawyer wrote: post #14
mejutoco wrote: post #4

Earlier quoted context omitted.

The key is if you consider weights source code. I do not think this is a common interpretation. > The labs themselves modify models the same as you are allowed to by the license Do the labs do not use source code? It is a bit like arguing that releasing a binary executable is releasing the source code. One could claim developers modify the binary the same as you are allowed to.

> Do the labs do not use source code? The weights are part of the source code. When running inference on a model you use the architecture, config files and weights together. All of these are released. Weights are nothing but "hardcoded values". The way you reach those values is irrelevant in the license discussion. Let's take a simple example: I write a chess program that is comprised of a source file with 10 "if" st…

> The weights are part of the source code.

If you will allow me the absurd analogy: my arm is also part of a person (me), but my arm is not a person. My arm does not have its own bank account and pays taxes independently.

I get how the weights are not exactly like binary code. Good points. But they are also not source code (from your own quote)

> "Source" form shall mean the preferred form for making modifications

The weights are not the preferred form of making modifications. At most, one could argue it is the weights + source code.

> In other words, there isn't another level above weights that the labs use to "compile" the weights

The source and training data?

I see your points, and it is an interesting discussion of nuances, but I profoundly disagree that the weights are "the preferred form for making modifications". For these reasons I prefer the term "open weights" for these projects.

Re: OpenAI's new open-source model is basically Phi-5

#136

Post by eddythompson80 »

numpad0 wrote: post #128
wmf wrote: post #20

I saw a bunch of people complaining on Twitter about how GPT-OSS can't be customized or has no soul and I noticed that none of them said what they were trying to accomplish. "The main use-case for fine-tuning small language models is for erotic role-play, and there’s a serious demand." Ah.

Maybe I'm just not seeing it, but is that use case really real and not just prudish hallucinations? The market for NSFW novels is smaller than even cyberpunk paperbacks, there's no way everyday people build up addiction for an interactive version of it.

Not NSFW, but I was surprised at hearing 2 random 20-something adults tell me they use chatgpt to write fan fiction to read. Those 2 people didn't know each other but were from similar demographic. It was surprising nonetheless. I can't imagine why anyone would care about such a bizarre use case.

Re: OpenAI's new open-source model is basically Phi-5

#137

Post by magicalhippo »

VladVladikoff wrote: post #18
magicalhippo wrote: post #15

I've found good use of Phi-4 at home, and after a few tests of the GPT-OSS 20B version I'm quite impressed so far. Particularly one SQL question that has tripped every other model of similar or smaller size that I've tried, like Devstral 24B, Falcon 3 7B, Qwen2.5-coder 14B and Phi 4 14B. The question contains an key point which is obvious for most humans, and which all of the models I tried previously have failed to…

Can you share the question? Or are you intentionally trying to keep it out of the training data pool?

Here's a more concrete example where GPT-OSS 20B performed very well IMHO. I tested it against Gemma 3 12B, Phi 4 Reasoning 14B, Qwen 2.5-coder 14B.

The prompt is modeled as a part of an agent of sorts, and the "human" question is intentionally ill-posed to emulate people saying the wrong thing.

The prompt begins with asking the model to convert a question into matlab code, add any assumptions as comments at the start of the coder, or if it's not possible then output four hash marks followed by an reason why.

The (ill-posed) question is "What's the cutoff frequency for an LC circuit with R equals 500 ohm and C equals 10 nanofarrad?"

Gemma 3 took the bait and treated R as L and proceeded to calculate the cutoff frequency of an LC circuit[1], completely ignoring the resulting mismatch of units. It did not comment at all. Completely wrong answer.

Qwen 2.5-coder detected the ill-posed nature, but instead decided to substitute a dummy value for L before calculating the LC circuit answer. On the upside it did add the comments saying this, so acceptable in that regard.

Phi 4 Reasoning reasoned for about 3 minutes before deciding to assume the question is about an RC circuit. It added this as a comment, and correctly generated the code for an RC circuit. So good answer, but slow.

GPT-OSS reasoned for 14 seconds, and determined the question was ill posed, thus outputting the hash marks followed by The cutoff frequency of an LC circuit cannot be determined with only R and C provided; the inductance L is required. Good answer, and fast.

[1]: https://en.wikipedia.org/wiki/LC_circuit#Resonance_effect

Re: OpenAI's new open-source model is basically Phi-5

#138

Post by seydor »

wmf wrote: post #20

I saw a bunch of people complaining on Twitter about how GPT-OSS can't be customized or has no soul and I noticed that none of them said what they were trying to accomplish. "The main use-case for fine-tuning small language models is for erotic role-play, and there’s a serious demand." Ah.

Why would a language company censor language like that? I literally don't see the purpose. Plenty of the best novels have explicit scenes. I certainly don't like novels about infantilized adults. Search engines don't do that, so why do we allow AI companies to do that so blatantly. It's basically culture engineering

Re: OpenAI's new open-source model is basically Phi-5

#139

Post by xwolfi »

eru wrote: post #55
LeoPanthera wrote: post #13

Earlier quoted context omitted.

By definition, a model can't "know" things that are not somewhere in its training set, unless it can use a tool to query external knowledge. The problem is that the size of the training set required for a good model is so large, that's really hard to make a good model without including almost all known written text available.

> By definition, a model can't "know" things that are not somewhere in its training set, unless it can use a tool to query external knowledge. Well, it could also make inferences. Like, it could find a new mathematical proof, even if that's never in the training set.

But how, it's not like it's thinking, it's just spitting the next likely token

Re: OpenAI's new open-source model is basically Phi-5

#140

Post by iamnotagenius »

magicalhippo wrote: post #137
VladVladikoff wrote: post #18

Earlier quoted context omitted.

Can you share the question? Or are you intentionally trying to keep it out of the training data pool?

Here's a more concrete example where GPT-OSS 20B performed very well IMHO. I tested it against Gemma 3 12B, Phi 4 Reasoning 14B, Qwen 2.5-coder 14B. The prompt is modeled as a part of an agent of sorts, and the "human" question is intentionally ill-posed to emulate people saying the wrong thing. The prompt begins with asking the model to convert a question into matlab code, add any assumptions as comments at the star…

glm 4 (1 sec):

To determine the cutoff frequency (fc ) for an RC circuit (since you've provided resistance R and capacitance C, but not inductance L), we can use the following formula:

[.... calculation]

So, the cutoff frequency is approximately 31.83 kHz.

Note:

If you intended to ask about an RLC circuit (with both R, L, and C), please provide the inductance L value, and I can calculate the cutoff frequency for that case as well. The formula would then involve both L and C.

Post reply on HN