msp26

12 ms·

151.
by msp26 2y ago
What are people's experiences with function calling on Claude? Does it work well for anything reasonably big?
152.
by msp26 2y ago
I still wanted the browser for UBlock Origin and handling sites with heavy JS. I was using the standalone Readability script already but today I ended up dropping it for Trafilatura. It works a lot better. The inefficiency of using a browse
153.
by msp26 2y ago
Great post. This exact limitation of web LLMs is why I'm leaning strongly towards local models for the easier stuff. Prompt caching can dramatically speed up fixed tasks. But frontier models are just too damn good and convenient so I d
154.
by msp26 2y ago
Thanks for the links I had no idea those existed. For my article web scraper (wip) the current steps are: - Navigate with playwright + adblocker - Run mozilla's readability on the page - LLM checks readability output If check failed -
155.
by msp26 2y ago
Pastebin? I don't really want to post my personal email on this account.
156.
by msp26 2y ago
If you show your task/prompt with an example I'll see if I can fix it and explain my steps. Are you using the function calling/tool use API?
157.
by msp26 2y ago
> But the problem is even worse – we often ask GPT to give us back a list of JSON objects. Nothing complicated mind you: think, an array list of json tasks, where each task has a name and a label. > GPT really cannot give back more th
158.
by msp26 2y ago
Absolutely
159.
by msp26 2y ago
Maybe the generated text could be a slightly different colour until it's verified. But you'd have to make sure there's no easy way of verifying everything mindlessly without having read it.
160.
by msp26 2y ago
No, you're along the right lines. Every prompting wrapper I've tried and looked through has been awful. It's not really the authors' faults, it's just a weird new problem with lots of unknowns. It's hard to get
161.
by msp26 3y ago
Oh so you work for newsapi?
162.
by msp26 3y ago
Do you have a contact email address? I'd love to talk more about this in detail. I've been working in the same space.
163.
by msp26 3y ago
> That is, assuming you're introverted enough that zero human interaction doesn't affect your mental health. This is never true. It might even feel this way for years but one day it will become unbearable.
164.
by msp26 3y ago
Friendly notice for anyone optimising their openai token usage: multi_tool use wastes 200 tokens on every call with function calling. You're better off implementing function calling yourself than using their API implementation.
165.
by msp26 3y ago
Pretraining and even finetuning (to a good extent) is overrated and you can create plenty of value without it.
166.
by msp26 3y ago
It's going to be amazing for hobbyists to get GPUs on the cheap.
167.
by msp26 3y ago
>After spending a lot of time with language models, I have come to the conclusion that tokenization in general is insane and it is a miracle that language models learn anything at all. Truer words have never been spoken
168.
by msp26 3y ago
Thank you for linking that paper!
169.
by msp26 3y ago
Hey, random question. Is there a technical reason why log probs aren't available when using function calling? It's not a problem, I've already found a workaround. I was just curious haha. In general I feel like the function c
170.
by msp26 3y ago
One of the constraints ended up being implemented into the main game: separate couriers for each player. But generally, agree with your point. But it's very cool how the OpenAI matches ended up making mid players reevaluate how they us
171.
by msp26 3y ago
So what is this actually putting into the prompt to guide generation? I dislike libraries that come with a lot of pointless abstraction. I'm about to write something that generates typescript code from pydantic models. If this just wor
172.
by msp26 3y ago
nono, the same thing happens with OpenAI's JSON mode and function calling. It does output parsable JSON very reliably but it comes out mangled with a bunch of whitespace sometimes. GPT-4-turbo's output context window is limited to
173.
by msp26 3y ago
Tokenization errors are also really, really annoying to deal with. e.g. why does the JSON output have silly whitespace/quotation sometimes? Obviously it's because the first token the model output was `{` and not `{"` like it
174.
by msp26 3y ago
Yeah that's definitely a risk with language models but it doesn't seem to be too bad for my use cases. Can I ask what tasks you used it for? I don't really intend for this method to be final. I'll switch everything over
175.
by msp26 3y ago
> I see plenty of use cases for such a big context, but re-paying, at every API call, to re-submit the exact same knowledge base seems very inefficient. If you don't care about latency or can wait to set up a batch of inputs in one
176.
by msp26 3y ago
It's amazing how much they've contributed to imagegen. I started using forge recently and it's a great speedup from regular sd-webui. https://github.com/lllyasviel/stable-diffusion-webui-forge
177.
by msp26 3y ago
I want image generators to generate what I ask them and not alter my query into something else. It's deeply shameful that billions of dollars and the hard work of incredibly smart people is mangled for a 'feature' that most e
178.
by msp26 3y ago
For me it was Ilya burning a wooden effigy that represented 'unaligned' AI. Of course the firing and twitter stuff too. Something's fucked in this company for sure.
179.
by msp26 3y ago
Surely someone can use a jailbreak to dump the context right? The same way we've been seeing how functions work.
180.
by msp26 3y ago
Unreal levels of shitposting. This whole model is high art.

More ›