Introduction
Introduction Statistics Contact Development Disclaimer Help
_______ __ _______
| | |.---.-..----.| |--..-----..----. | | |.-----..--.--.--..-----.
| || _ || __|| < | -__|| _| | || -__|| | | ||__ --|
|___|___||___._||____||__|__||_____||__| |__|____||_____||________||_____|
on Gopher (inofficial)
Visit Hacker News on the Web
COMMENT PAGE FOR:
Curious about the training data of OpenAI's new GPT-OSS models? I was too
ComputerGuru wrote 1 day ago:
Not very rigorous or scientific, honestly, I would say it's just
clickbait spam with some pretty graphs. Everything on twitter is now a
"deep dive". No info on how the 10M "random examples" were generated
and how that prevents the model from collapsing around variations of
the same output. Others already mentioned how the "classification" of
output by coding language is bunk with a good explanation for how Perl
can come out on top even if it's not actually Perl, but I was struck by
OP saying "(btw, from my analysis Java and Kotlin should be way higher.
classifier may have gone wrong)" but then merrily continuing to use the
data.
Personally, I expect more rigor from any analysis and would hold myself
to a higher standard. If I see anomalous output at a stage, I don't
think "hmm looks like one particular case may be bad but the rest is
fine" but rather "something must have gone wrong and the entire
output/methodology is unusable garbage" until I figure out exactly how
and why it went wrong. And 99 times out of a 100 it wasn't the one case
(that happened to be languages OP was familiar with) but rather
something fundamentally incorrect in the approach that means the data
isn't usable and doesn't tell you anything.
godelski wrote 23 hours 34 min ago:
> Personally, I expect more rigor from any analysis and would hold
myself to a higher standard.
When something is "pretty bizarre" the most likely conclusion is "I
fucked up", which is very likely in this case. I really wonder if he
actually checked the results of the classifier. These things can be
wildly inaccurate since differences in languages can be quite small
at times and some are very human language oriented. He even admits
that Java and Kotlin should be higher but then doesn't question Perl,
R, Applescript, Rust, and the big drop to Python. What's the joke? If
you slam your head on the keyboard you'll generate a valid Perl
program?
It worries me that I get this feeling from quite a number of ML
people who are being hired and paid big bucks from big tech
companies. I say this as someone in ML too. There's a propensity to
just accept outputs rather than question them. This is like a basic
part of doing any research, you should always be incredibly
suspicious of your own results. What did Feynman say? Something like
"The first rule is not to be fooled and you're the easiest person to
fool"?
greenchair wrote 1 day ago:
"this thing is clearly trained via RL to think and solve tasks for
specific reasoning benchmarks. nothing else." Has the train already
reached the end of the line?
red75prime wrote 15 hours 30 min ago:
If you think something like "They have to train their models on
benchmarks to make it look like there's progress, while in reality
it's a dead end," you are missing a few things.
It's an open model, everyone can bench it on everything not only on
specific benchmarks. Training on specific reasoning benchmarks is a
conjecture.
jdfr wrote 1 day ago:
OP seems to have run a programming language detector on the generated
texts, and made a graph of programming language frecuencies: [1] As a
result, OP seems to think the model was trained on a lot of Perl: [2]
LOL! I think these results speak more to the flexibility of Perl than
any actual insight on the training data! After all, 93% of inkblots are
valid Perl scripts:
[1]: https://pbs.twimg.com/media/Gx2kvNxXEAAkBO0.jpg?name=orig
[2]: https://xcancel.com/jxmnop/status/1953899440315527273#m
[3]: https://www.mcmillen.dev/sigbovik/
godelski wrote 23 hours 48 min ago:
Jack has a lot of really bad takes and frequently makes lots of
mistakes. Honestly, I don't know why people take him seriously.
I mean you can go read his blog post that's pinned where he argues
that there's no new ideas and it is all data. He makes the argument
that architecture doesn't matter, which is just so demonstrably false
that it is laughable. He's a scale maximalist.
I also expect an AI researcher from a top university to not make such
wild mistakes
> 3. RLHF: first proposed (to my knowledge) in the InstructGPT
paper from OpenAI in 2022
I mean if you go read the instruct paper on page 2 you'll see
| Specifically, we use reinforcement learning from human feedback
(RLHF; Christiano et al., 2017; Stiennon et al., 2020) to fine-tune
GPT-3 to follow a broad class of written instructions (see Figure 2).
Where in Christiano you'll find
|| Our algorithm follows the same basic approach as Akrour et al.
(2012) and Akrour et al. (2014)
I mean this is just obviously wrong. It is so obviously wrong it
should make the person saying it second guess themselves (which is
categorically the same error you're pointing out).
I'm sure we can trace the idea back to the 80's if not earlier. This
is the kind of take I'd expect a non-researcher to have, but not
someone with two dozen NLP papers. The Instruct-GPT paper was just
the first time someone integrated RLHF into a LLM (but not a LM).
Maybe a better article is the one he wrote on Super Intelligence From
First Principles. As usual, when someone says "First Principles" you
bet they're not gonna start from First Principles... I guess this
makes sense in CS since we index from 0
[0] [1] [Christiano et al] [2] [Stiennon eta al] [3] [Akrour et al
(2012)]
[1]: https://arxiv.org/abs/2203.02155
[2]: https://arxiv.org/abs/1706.03741
[3]: https://arxiv.org/abs/2009.01325
[4]: https://arxiv.org/abs/1208.0984
jxmorris12 wrote 21 hours 29 min ago:
Hi again. I had already written about this later in my blog post
(which is unrelated to this thread), but the point was that RLHF
hadn't been applied to language models at scale until InstructGPT.
I edited the post just now to clarify this. Thanks for the
feedback!
actuallyalys wrote 1 day ago:
Honestly these results may say as much about the classifier as they
do about the data they’re classifying.
esafak wrote 1 day ago:
I don't understand why Perl, R, and AppleScript rank so much higher
than their observed use.
jxmorris12 wrote 21 hours 28 min ago:
It seems to be an error with the classifier. Sorry everyone. I
probably shouldn't have posted that graph; I knew it was buggy, I
just thought that the Perl part might be interesting to people.
Here's a link to the model if you want to dive deeper:
[1]: https://huggingface.co/philomath-1209/programming-language...
j_bum wrote 1 day ago:
R being so high makes no sense to me either.
I think as of the last Stack Overflow developer survey, it only had
~4% market share…
I say this as an R user who spams LLMs with R on a daily basis.
londonlawyer wrote 1 day ago:
The prominence of AppleScript ought to have been a pretty big red
flag: the author seems to be claiming the model was trained on more
AppleScript than Python, which simply can’t be true.
Ironically LLMs seem pretty bad at writing AppleScript, I think
because (i) the syntax is English-like but very brittle, (ii) the
application dictionaries are essential but generally not on the
web, and (iii) most of the AppleScript that is on the web has been
written by end users, often badly.
rozab wrote 1 day ago:
Perl and Applescript are close to natural language. R is close to
plain maths
[1]: https://en.wikipedia.org/wiki/Black_Perl
johnisgood wrote 1 day ago:
That inkblot thing can be created for any language.
westurner wrote 1 day ago:
Of the powerset of all operators and inputs, how many can be
represented in any programming language?
What percent of all e.g. ASCII or Unicode strings are valid
expressions given a formal grammar?
bravesoul2 wrote 1 day ago:
How? E.g. I doubt an inkblot can produce a valid C# program.
johnisgood wrote 1 day ago:
They are not full programs, just code translating to numbers and
strings.
I used an LLM to generate an inkblot that translates to a Python
string and number along with verification of it, which just
proves that it is possible.
bstsb wrote 1 day ago:
what are you talking about?
the way that the quoted article creates Perl programs is
through OCRing the inkblots (i.e. creating almost random text)
and then checking that result to see if said text is valid Perl
it's not generating a program that means anything
johnisgood wrote 1 day ago:
Okay, and I created inkblots that mean "numbers"[1] and
"strings" in Python.
> it's not generating a program that means anything
Glad we agree.
[1] Could OCR those inkblots (i.e. they are almost random
text)
mathiaspoint wrote 1 day ago:
Most random unquoted strings are certainly not valid Python
programs. I don't know Perl well enough to say anything
about that but I know what you're saying certainly isn't
true with Python.
dmbche wrote 1 day ago:
No, asking an LLM to generate the inkblot is the same as
asking the LLM to write a string and then obfuscating it in
an inkblot.
OCRing literal random inkblots will not produce valid C (or
C# or python) code, but it will prodce valid Perl most of
the time, because Perl is weird, and that is funny.
It's not about obfuscating text in inkblot, it's about
almost any string being a valid Perl program, which is not
the case for most languages
Edit0: here:
[1]: https://www.mcmillen.dev/sigbovik/
johnisgood wrote 1 day ago:
Okay, my bad.
> it's about almost any string being a valid Perl program
Is this true? I think most random unquoted strings aren't
valid Perl programs either, am I wrong?
akerl_ wrote 1 day ago:
Yes. That was the whole point of the original comment
you were misunderstanding.
Because of the flexibility of Perl and heavy amount of
symbol usage, you can in fact run most random
combinations of strings and they’ll be valid Perl.
Copying from the original comment:
[1]: https://www.mcmillen.dev/sigbovik/
ma2rten wrote 1 day ago:
Presumably the model is trained in post-training to produce a response
to a prompt, but not to reproduce the prompt itself. So if you prompt
it with an empty prompt it's going to be out of distribution.
james-bcn wrote 1 day ago:
This looks very interesting but I don't really understand what he has
done here. Can someone explain the process he has gone through in this
analysis?
AmazingTurtle wrote 1 day ago:
He presented an empty prompt to gpt OSS and let it run many times.
Through temperature, the results vary quite a lot. He sampled the
results.
Feeding an empty prompt to a model can be quite revealing on what
data it was trained on
YeGoblynQueenne wrote 1 day ago:
Not an empty prompt but a one-token prompt:
>> i sample tokens based on average frequency and prompt with 1
token
[1]: https://x.com/iamgrigorev/status/1953919577076683131
esperent wrote 1 day ago:
> the chains start in English but slowly descend into Neuralese
What is Nueralese? I tried searching for a definition but it just turns
up a bunch of Less Wrong and Medium articles that don't explain
anything.
Is it a technical term?
spwa4 wrote 1 day ago:
There's 2 things called neuralese:
1) internally, in latent space, LLMs use what is effectively a
language, but all the words are written on top of each other instead
of separately, and if you decode it as letters, it sounds like
gibberish, even though it isn't. It's just a much denser language
than any human language. This makes them unreadable ... and thus
"hides the intentions of the LLM", if you want to make it sound
dramatic and evil. But yeah, we don't know what the intermediate
thoughts of an LLM sound like.
The decoded version is often referred to as "neuralese".
2) if 2 LLMs with sufficiently similar latent space communicate with
each other (same model), it has often been observed that they switch
to "gibberish" BUT when tested they are clearly still passing
meaningful information to one another. One assumes they are using
tokens more efficiently to get the latent space information to a
specific point, rather than bothering with words (think of it like
this: the thoughts of an LLM are a 3d point (in reality 2000d, but
...). Every token/letter is a 3d vector (meaning you add them),
chosen so words add up to the thought that is their meaning. But when
outputting text why bother with words? You can reach any
thought/meaning by combining vectors, just find the letter moving the
most in the right direction. Much faster)
Btw: some specific humans (usually toddlers or children that are
related) when talking to each other switch to talking gibberish to
each other as well while communicating. This is especially often
observed in children that initially learn language together. Might be
the same thing.
These languages are called "neuralese".
bananaflag wrote 1 day ago:
[1]: https://en.wikipedia.org/wiki/Poto_and_Cabengo
nopinsight wrote 1 day ago:
The author might use it as an analogy to mentalese but for neural
networks. [1] EDIT: After reading the original thread in more detail,
I think some of the sibling comments are more accurate. In this
case, neuralese is more like language of communication expressed by
neural networks, rather than its internal representation.
[1]: https://en.wiktionary.org/wiki/mentalese
meowface wrote 1 day ago:
It's a term somewhat popularized by the LessWrong/rationalism
community to refer to communication
(self-communication/note-taking/state-tracking/reasoning, or
model-to-model communication) via abstract latent space information
rather than written human language. Vectors instead of words.
One implication leading to its popularity by LessWrong is the worry
that malicious AI agents might hide bad intent and actions by
communicating in a dense, indecipherable way while presenting only
normal intent and actions in their natural language output.
verisimi wrote 1 day ago:
> malicious AI agents might hide bad intent and actions by
communicating in a dense, indecipherable way while presenting only
normal intent and actions in their natural language output.
you could edit this slightly to extract a pretty decent rule for
governance, like so:
> malicious agents might hide bad intent and actions by
communicating in a dense, indecipherable way while presenting only
normal intent and actions in a natural way
It applies to ai, but also many other circumstances where the
intention is that you are governed - eg medical, legal, financial.
Thanks!
ben_w wrote 1 day ago:
Easier said than done:
• [1] • [2] Or even just regional differences, like how
British people, upon hearing about "gravy and biscuits" for the
first time, think this: [3] > It applies to ai, but also many
other circumstances where the intention is that you are governed
- eg medical, legal, financial.
May be impossible to avoid in any practical sense, due to every
speciality having its own jargon. Imagine web developers having
to constantly explain why "child element" has nothing to do with
offspring.
[1]: https://en.wikipedia.org/wiki/Cant_(language)
[2]: https://en.wikipedia.org/wiki/Dog_whistle_(politics)
[3]: https://thebigandthesmall.com/blog/2019/02/26/biscuits-g...
fl1pper wrote 1 day ago:
neuralese is a term first used in neuroscience to describe the
internal coding or communication system within neural systems.
it originally referred to the idea that neural signals might form an
intrinsic "language" representing aspects of the world, though these
signals gain meaning only through interpretation in context.
in artificial intelligence, the term now has a more concrete role,
referring to the deep communication protocols used by multiagent
systems.
CjHuber wrote 1 day ago:
I suppose it means LLM gibberish
EDIT: orbital decay explained it pretty well in this thread
puttycat wrote 1 day ago:
> OpenAI has figured out RL. the models no longer speak english
What does this mean?
tehnub wrote 1 day ago:
I think foremost it's a reference to this tweet [1] .
[1]: https://x.com/karpathy/status/1835561952258723930
orbital-decay wrote 1 day ago:
The model learns to reason on its own. If you only reward correct
results but not readable reasoning, it will find its own way to
reason that is not necessarily readable by a human. The chain may
look like English, but the meaning of those words might be completely
different (or even the opposite) for the model. Or it might look like
a mix of languages, or just some gibberish - for you, but not for the
model. Many models write one thing in the reasoning chain and a
completely different in the reply.
That's the nature of reinforcement learning and any evolutionary
processes. That's why the chain of thought in reasoning models is
much less useful for debugging than it seems, even if the chain was
guided by the reward model or finetuning.
Hard_Space wrote 1 day ago:
Interesting. This happens in Colossus: The Forbin Project (1970),
where the rogue AI escapes the semantic drudgery of English and
invents its own compressed language with which to talk to its Russian
counterpart.
Mistletoe wrote 1 day ago:
It also happens in Ex Machina at the end when the two androids
whisper and talk to each other in their special faster language. I
always found this to be one of the most believable, real things
from that movie and one of my favorite parts.
pinoy420 wrote 1 day ago:
5 seems to do a better job with copyrighted content. I got it to spit
out the entirely of ep IV (but you have to redact the character names)
revskill wrote 1 day ago:
What does that mean ?
orbital-decay wrote 1 day ago:
>what you can't see from the map is many of the chains start in English
but slowly descend into Neuralese
That's just natural reward hacking when you have no
training/constraints for readability. IIRC R1 Zero is like that too,
they retrained it with a bit of SFT to keep it readable and called it
R1. Hallucinating training examples if you break the format or prompt
it with nothing is also pretty standard behavior.
k310 wrote 2 days ago:
Anything but this image (imgbb.com link below) requires a login. I get
the same deal with Facebook. I am not Don Quixote and prefer not to
march into hell for a heavenly cause, nor any other.
[1]: https://i.ibb.co/Zz2VgY4C/Gx2-Vd6-DW4-AAogtn.jpg
Epskampie wrote 1 day ago:
[1]: https://xcancel.com/jxmnop/status/1953899426075816164
k310 wrote 1 day ago:
Thanks! I've seen a lot of stuff come and go, so thanks for the
reminder.
For example, Libgen is out of commission, and the substitutes are
hell to use.
Summary of what's up and not up:
[1]: https://open-slum.org/
stavros wrote 1 day ago:
Oh no, why did Libgen die?
k310 wrote 1 day ago:
Shut down. See [1] for substitutes. The alternate libgen sites
seem more limited to me, but I am comparing with memories, so
untrustworthy.
[1]: https://open-slum.org/
nikcub wrote 1 day ago:
it's available at the bz tld
randomNumber7 wrote 1 day ago:
> Libgen is out of commission, and the substitutes are hell to
use
Somehow I also preferred libgen, but I don't think annas archive
is "hell to use".
k310 wrote 1 day ago:
Annas Archive uses slow servers on delay, and constantly tells
me that they are too many downloads from my IP address, so I
flip VPN settings as soon as the most recent slow download
completes. And I get it again after a short while. It's hell
waiting it out and flipping VPN settings. And the weird part is
that this project is to replace paper books that I already
bought. That's the excuse one LLM uses for tearing up books,
scanning and harvesting. I just need to downsize so I can move
back to the Bay Area. Book and excess houseware sale coming,
it seems. Libgen had few or no limits.
1gn15 wrote 1 day ago:
I would recommend donating to gain access to the fast
downloads; they need money for the servers.
flabber wrote 2 days ago:
I don't know how to get a unwalled version. What's the best way to do
that these days? xcancel seems unavailable.
mac-attack wrote 1 day ago:
Install libredirect extension ( [1] ) and select a few working
instances. Then you can use the programmable shortcut keys to cycle
between instances if one ever goes down.
[1]: https://github.com/libredirect/browser_extension/
striking wrote 1 day ago:
xcancel is fine, here's an archive of it:
[1]: https://archive.is/VeUXH
k310 wrote 1 day ago:
Thanks!
<- back to front page
You are viewing proxied material from codevoid.de. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.