Hacker News new | past | comments | ask | show | jobs | submit | bl4ckneon's comments login

Just thinking that. Spend a few minutes trying to have chatgpt generate some images with Dall-E 3. Flux would probably be better to get all the specific details but ya

(didn't look at the code yet but) Would a challenge of building an app like this that heavily depends on a LLM be getting a deterministic response back? I guess you could code for it to check if it gave you a certain format of data or if it was what you expected, but if I upload something that Claude doesn't understand and it gives back something that breaks the data analysis then that seems it would be tricky to handle that case.

Please correct me if I am wrong. Thanks!


Anthropic and OpenAI let you define a JSON schema to adhere to for tool calling.

Here's the part you're looking for: https://github.com/anthropics/anthropic-quickstarts/blob/mai...


For some reason, the guarantee in the format of the response doesn't seem sufficient in preventing backwards incompatible changes that may happen to models.

Yes, the response might be in a standard format. But a well formed response can still be bad/broken.

Another way to think about it, is it can "pass QA" one day, and "fail QA" the next day even if the API response is identically formatted/structured.


This is why OpenAI and Anthropic provide date versioned models.

gpt-4o can change, but gpt-4o-2024-05-13 will always use the 2024-05-13 snapshot.


i have a feeling those dates are an illusion of sorts.

I get the feeling they frequently deploy hot patches for edge cases. I hate to call them edge cases because they are actually “real cases” - things like adjusting system prompts so one day it might happy answer “Fill in the blank: F _ _ _ you”.

To truly freeze a model, you would need to freeze its weights, freeze its system prompts (no one sees those), and avoid any and all action that might impact its output. Perhaps would even need the default temperature to be 0 so it’s truly a deterministic API, with the option to add in some temperature to the responses.

Until then, I consider those “versions” but only reference the model weights and not the abstractions around the model


Tangent, but it seems like such a tough engineering challenge to keep all these models around and available at an instant


My cat often lays down twisted 180 degrees or more. Just doing whatever they want, defying laws of nature.


Well, dogs also do this—I present to you my majestically twisted creature: https://imgur.com/a/5WcYzSw

I have no clue how that is even possible.


Brought memories of one of my cats (now silent meow) who also added the Italian equivalent of a middle finger.

https://imgur.com/a/GFukfFP



Yeah, nope. If I get like that, I'm never coming back. Probably have to bury me in that pose.

Is this really just a matter of stretching? I read the article and he sums it down to he needs to stretch every day (he said himself thst his diet doesn't matter too much) He was also in the circus since 4, but this doesn't seems like something I could do in a lifetime of practice.


You probably couldn't. There are lots of forms of hyper mobility, and extreme versions come with health risks. With practice and training you can probably do a lot more than you imagine, but for most of us the whole "fold yourself in half backwards" thing is beyond the limits of our spine, and it's for the best.


No, it's not just down to stretching although I remember being told that as a kid when asking about "snake ladies" in circuses. Yes, the stretching and training is necessary but all the stretching and training in the world won't help you bend like this. Some people simply have hypermobilities (often from connective tissue disorders) and these can allow them to bend themselves into a pretzel. The stretching and training is as much for learning how to take it to such extremes as it is for learning how to do it (relatively) safely. Of course these conditions can still often lead to complications and often have other negative impacts like chronic pain, frequent bruising, slower healing and a tendency to dislocate various joints.


Your dog is the inverse of the Firefox.


My Shiba Inu does all kinds of similar things. He also doesn't hesitate at all when trying narrow spaces. He only hesitates once he's all the way in and realize he can't go any further nor turn around so he has to back up completely back out.


I almost sprayed all my tea to my monitor and keyboard.

Wish both of you a happy and derpy life together.


Mine managed to fit into this tiny basket, all by himself: https://i.imgur.com/htIw6ZV.jpeg


Clearly your dog has been possessed by a demon.


majestic indeed!


Dog Yoga


Doga.


    $ meme init
    meme template initiated


That's actually not true. At a certain size of company and especially if the company does work for the federal government they are required to gather demographic information to ensure they are trying to not just have a bunch of white dudes applying for their job. Those questions should not influence the hiring process at all (if they do then it's illegal since those are protected traits).

If you don't want to answer them then just answer it by stating you don't want to tell them. It's an option on every application I applied too.


Take a look at https://github.com/Stirling-Tools/Stirling-PDF

This might be the tool you are looking for. Locally hosted too


Thanks. Very much what I'm looking for.


Self hosted LLM?


I had that too, no explanation, no response via email support, nothing. I can't give money to them knowing my account can just get banned at anytime while I might have an active subscription going on.


It's also intresting to see a language used in a way it's not intended to necessarily our for a creative/unique use. Why do people like to port doom to random devices, because it was never indended to run on those (and it's a challenge/meme at this point)


I believe it was a bunch of them in a trenchcoat!


Do we have proof of that though. Other than "China bad" and the parent company being partially owned by China. It's really banning it on a hypothetical and TikTok doesn't exist in a vacuum. There are many other platforms and I would argue that X does more censorship than TikTok does.


Yeah, I think there is lots evidence that social media companies use their platform to support favourable narratives. X is one example.

X spinning a pro X narrative is problematic, just as it is bad Facebook hid their problem with teenage mental health, etc... all these types of things are problematic and can often harm people and society.

The reason TikTok is a greater concern is that TikTok is fully controlled by China, and that hence it is not just pro TikTok narratives that might be amplified but any narrative China wants to amplify. The risks are greater, and the existing mitigations are weaker as the company isn't US owned.

If you ask me, the US in recent history has had incredibly anemic corporate regulation, and there should be a lot more intervention to prevent companies acting nefariously.


Just to be clear, your position is that it could be used by China, not that it is being used? I agree with this position.


And that's sufficiently concerning to nip this in the bud now. It is either happening or could be made to happen with ease and neither of those are good outcomes.


Consider applying for YC's Spring batch! Applications are open till Feb 11.

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

Search: