HNHacker News
TopNewBestAskShowJobs

radarsat1

4,157 karma · joined May 28, 2011

Info: https://sinclairs.gitlab.io

Email: <myusername>@gmail.com

submissionscomments
radarsat1··on Detecting when LLMs are uncertain
Sampling sequentially to find the highest joint probability over the sequence is definitely a search problem. that's why you see algorithms like beam search often used for sampling.
radarsat1··on Detecting when LLMs are uncertain
The problem is that deep net classifiers in general are not well statistically calibrated by default. So while the entropy is often high when they are "not sure", models can very often also be "confidently wrong". So using entropy of the logits as an indicator of confidence can easily be very misleading.

I'm not an expert in LLMs though, this is just my understanding of classifiers in general. Maybe with enough data this consideration no longer applies? I'd be interested to know.

radarsat1··on Adding syntax to the CPython interpreter
Why would I want to extend the language syntax as part of normal execution?
radarsat1··on The science of "Zoom fatigue"
So, so many of the issues discussed in this thread are related to video. I've found that video is almost not really necessary. Unfortunately many people expect it these days, but if you can get some consensus that "we don't need to turn on cameras for this" then I find things go just as smoothly, if not moreso.

It's really a shame that people have made it almost weird to not have cameras on. Admittedly it can be nice to see the other participants in the meeting but I find it's only useful for the first 5 minutes, then it may as well go to audio. Saves on bandwidth too.

radarsat1··on Show HN: Automated smooth Nth order derivatives of noisy data
Did you try prefiltering to remove the repeated values?
radarsat1··on INTELLECT–1: Launching the First Decentralized Training of a 10B Parameter Model
> 1 bit per weight

does this basically correspond to moving each weight either up or down by a fixed amount? I'm a bit surprised you don't at least need a "stay same" bit, but i suppose it could balance out over multiple iterations.

Interesting that it works at all. Although, thinking on it, I could see it maybe even having a nice regularizing effect where every layer would end up have similar weight magnitudes. (like projecting onto the local n-ball as mentioned in a paper posted recently on HN)

radarsat1··on Differential Transformer
Agreed. I have a loose idea that hallucination is related to training to maximize the probability of individual tokens while ignoring the joint sequence probability, which is along the lines of what you are saying -- it is not trained to output the most probable final sequence, so it gets stuck in the "wrong place" half way through.
radarsat1··on Differential Transformer
Often see this argument but it doesn't hold water for me. What we call hallucination is usually when the model says something confidently wrong. Yes the sampling procedure is nondeterministic but this is unrelated to hallucinations. The model can generate a distribution to sample with very little weight on the "wrong" output and then this should be ignored by procedures like top-k sampling. The fact that this doesn't easily solve the problem shows that hallucination is a deeper problem in the model itself and not just a byproduct of sampling.
radarsat1··on Were RNNs all we needed?
Interstellar taught me that love transcends the universe. Ergo..
radarsat1··on Show HN: A real time AI video agent with under 1 second of latency
As someone not super familiar with deployment but enough to know that GPUs are difficult to work with due to being costly and sometimes hard to allocate: apart from optimizing the models themselves, what's the trick for handling cloud GPU resources at scale to serve something like this, supporting many realtime connections with low latency? Do you just allocate a GPU per websocket connection? Which would mean keeping a pool of GPU instances allocated in case someone connects, otherwise cold start time would be bad.. but isn't that super expensive? I feel like I'm missing some trick in the cloud space that makes this kind of thing possible and affordable.
radarsat1··on Twitter shut off API access; users volunteering their own data for an open API
> Do people, like, realize that their whole process for protecting scraping the social network via third-party app integration is the honor system?

That sentiment goes back pretty far in Facebook's history

[1] https://www.yahoo.com/news/mark-zuckerberg-branded-early-fac...

radarsat1··on JG Ballard's Apocalyptic Art
Nice thanks for that, it's been a while since both I read the novel and saw the film. That's actually a much more nuanced and intriguing take. I might have to reread it now! This article makes me want to read the rest of his works, too.
radarsat1··on JG Ballard's Apocalyptic Art
I thought the point was more that although everyone in the building were wealthy, some were more wealthy than others, a comment on the human nature of letting even the most miniscule differences take on adversarial meaning when isolated.
radarsat1··on GPTs and Hallucination
I think this may be the best explanation I've seen on the topic!

But, shouldn't that situation be handled somewhat by backtracking sampling techniques like beam search? But maybe that is not used much in practice due to being more expensive.. don't know.

radarsat1··on GPTs and Hallucination
what you're referring to has nothing to do with how GPTs are pretrained or with hallucinations in and of themselves, and everything to do with how companies have reacted to the presence of hallucinations and general bad behavior, using a combination of fine tuning, RLHF, and keyword/phrase/pattern matching to "guide" the model and cut it off before it says something the company would regret (for a variety of reasons)

In other words, your complaints are ironically not about what the article is discussing, but about, for better or for worse, attempts to solve it.

radarsat1··on The Internet Archive has lost its appeal in Hachette vs. Internet Archive
What will this mean for other sites with similar programs? I know many libraries do similar "renting" of digital texts on their websites. I guess they will have to shut down these programs.

For instance, I know this is Canadian so different jurisdiction but here is an example of one: https://montreal.ca/en/services/digital-book-loans

radarsat1··on Diffusion models are real-time game engines
Most games are conditioned on text, it's just that we call it "source code" :).

(Jk of course I know what you mean, but you can seriously see text prompts as compressed forms of programming that leverage the model's prior knowledge)

radarsat1··on Diffusion models are real-time game engines
There has definitely been research for simulating physics based on observation, especially in fluid dynamics but also for rigid body motion and collision. It's important for robotics applications actually. You can bet people will be applying this technique in those contexts.

I think for real world application one challenge is going to be the "action" signal which is a necessary component of the conditioning signal that makes the simulation reactive. In video games you can just record the buttons, but for real world scenarios you need difficult and intrusive sensor setups for recording force signals.

(Again for robotics though maybe it's enough to record the motor commands, just that you can't easily record the "motor commands" for humans, for example)

radarsat1··on Show HN: Attaching to a virtual GPU over TCP
Aah ok thanks, that was my basic misunderstanding, my mind just jumped straight to my current training needs but for inference it makes a lot of sense. Thanks for the clarification.
radarsat1··on Show HN: Attaching to a virtual GPU over TCP
I'm confused, if this operates at the CPU/GPU boundary doesn't it create a massive I/O bottleneck for any dataset that doesn't fit into VRAM? I'm probably misunderstanding how it works but if it intercepts GPU i/o then it must stream your entire dataset on every epoch to a remote machine, which sounds wasteful, probably I'm not getting this right.
radarsat1··on What Is Analog Computing?
Isn't that where the term analog for continuous values comes from though? That they are voltages that vary in proportion to physical quantities? And therefore analogous to those quantities.
radarsat1··on Show HN: Datoviz – Vulkan-based GPU scientific visualization (C/C++/Python)
Is this similar in spirit to VTK? (An API I've not always found super intuitive so interested in alternatives.)

Or is this better thought of as a powerful matplotlib alternative?

radarsat1··on Structured Outputs in the API
Looks useful!
radarsat1··on The Path to StyleGan2 – Implementing the Progressive Growing GAN
Latent diffusion models operate in latent space. This space is generated by an encoder and decoded back into pixel space by a decoder. The encoder-decoder form a generator which is trained to have good visual quality through the use of an adversarial loss.

So the encoder produces a latent space that is more efficient to train a diffusion model on, since diffusion models use Unet-like architecture that must be run many times for a single inference. The latent space is restricted by a KL penalty to a Gaussian shape such that any sample from that shape will map through the decoder to a high-quality image. This makes the generative job of the diffusion model much easier because it can focus on content and semantics rather than pixel-level details

You can see the two optimisers at work in the AutoencoderKL class in the Stable Diffusion source code here: https://github.com/CompVis/stable-diffusion/blob/main/ldm/mo...

radarsat1··on The Path to StyleGan2 – Implementing the Progressive Growing GAN
> I didn’t know that GANs were still in use

Their still in use literally by every latent diffusion image generator, they typically target the latents of a GAN-trained decoder.

Same for audio, most audio models generate some representation that is converted to audio by a GAN-trained codec.

radarsat1··on RustPython: A Python Interpreter Written in Rust
Right so you're saying that Python's need to keep ref counts is what leads to the need for synchronizing updates, leading to the need for a lock, more or less. Which is only needed in C++ if you program in a kind of Python style. Makes sense and is a good point.
radarsat1··on IsoFLOP curves of large language models are flat
Thanks that's actually pretty much what I was thinking, I'll have to read that to try to understand the significance of the S curve there, pretty interesting. Appreciate the link!
radarsat1··on RustPython: A Python Interpreter Written in Rust
> yes, people do depend on objects being eagerly freed

I get that this must be one aspect of the necessity of the GIL but I mean, C++ also has eager free behavior due to RAII and threads are working fine there, as long as you know what you're doing. Perhaps that's the rub though, it's pretty easy to crash/deadlock in C++ and we blame the programmer rather than the language.

radarsat1··on IsoFLOP curves of large language models are flat
> So, these models are basically within 1% of each other in terms of final pretraining loss.

How is this loss calculated though? Since it is called "loss" and not "performance metric", I'm going to assume it is the teacher forced cross entropy loss.

I'm not too familiar with LLM training but having been doing some fair amount of seq2seq training lately in other domains I've observed that the relationship between "loss" and autoregressive inference performance gets very narrow towards the end of training. What I mean is that smaller and smaller reductions in loss lead to better improvements in the autoregressive output. So at least I have some doubt that in practice that 1% loss improvement is not actually incredibly significant with respect to how well the model actually performs at inference time.

But I'm pretty interested in this topic and if people here have observed this or the contrary I'd be curious to know.

radarsat1··on The oldest known recording of a human voice [video]
This is cool because I've heard the recording before but didn't know the story behind it.

What's funny to me is, looking at that invention design, how crazy this guy must have appeared to his peers, like, "look it writes the sound into the ashes!!!". "Sure Eddy, buddy, let's get you a nice cup of tea and calm down.."

Yet he was on to something amazing that would change how we live.

I suppose there was a "crazy inventor" culture at the time though, with so much new understanding of mechanical physics and engineering developing at a such a rapid pace, so maybe it wasn't so out of place, what a time that must have been to be alive..

← PreviousPage 6 of 34Next →