@jarredsumner I remember doing a few PRs on Bun, but the compile times were so long after each minor change that I ended up no longer contributing. Is this fixed with the new Rust rewrite?
@crungulism@HSVSphere Have you checkout out burn? It has no-std support for any onnx model. Which means you can import from pretty much any machine learning framework. Might be useful.
@LiveOverflow@ZetaTwo I would turn the comments into embeddings, then do a text similarity comparison for mentions of color.
Probably faster, cheaper, and more reliable than using GPT to moderate.
Also, at the point that you want to do content moderation, it would be worth training a custom model.
@imHitchens@LiveOverflow@ZetaTwo I managed to bypass this prompt, very easy to prevent if you added a word/token limit, but the way to bypass this is by adding a bunch of fake statements attributed to the @LiveOverflow.
Note that no new line characters were added to the @ZetaTwo user.
https://t.co/HQhRrDdUnw
we are starting our rollout of ChatGPT plugins.
you can install plugins to help with a wide variety of tasks. we are excited to see what developers create!
https://t.co/NQ684Yp2LK
@KevinAFischer Check out @AiEleuther’s. They have open source models, which are similar to OpenAI’s GPT.
I believe their largest model is GPT-NeoX-20B which is one click deploy on huggingface.
Definitely check this out as an alternative to giving all your data away.
https://t.co/uwvHc4Mkzk
@KevinAFischer I feel the same way, but this fundamentally doesn’t prevent someone from just writing a wrapper around the API and publicizing that.
So in the end you’re forced to make the same restrictions. Though I wonder if fine tuning can mitigate it.
Would be interesting to investigate.
@KevinAFischer Yes, I understand this viewpoint, and I’m constantly fighting with GPT to make it actually output something useful or remotely interesting, but it always falls flat (i.e. writing stories).
At the same time I get the rationale of OpenAI not wanting misuse of GPT (i.e. hacking).
@Zane97070545 Lol, I have a feeling that this post must be a troll. It seems just too unreal that someone would complain about their 9-year old not making production code.
@laceynwilliams This is completely insane, I remember when I was 8 I got full support from my brother to learn programming. Never would have learned everything I did/be so passionate without that.
Just imagine what that does to a kids prospective and ambitions.
Plus the undertones…
@Zane97070545 Should be required in every paper. You somehow have to incorporate one, and relevantly.
Nothing is funnier than when you’re reading a presumably bland textbook and then the author has some hilarious dry humor in it.
I laughed out loud upon reading the paper for GPT-4. They compared the accuracy of the different models of GPT using.... standardized tests!
https://t.co/wu2tqwAbob
🚨 HUGE news in AI: Google just launched Generative AI across ALL of Google Workspace -- Gmail, Docs, Sheets, Slides, Images -- EVERYTHING.
They made a video showing off the new AI's capabilities. It's AWESOME.
Nice to see GPT-4 was able to solve many more Leetcode problems successfully, I wonder if it will address some of the issues in this study I read a few months back, demonstrating that users write more insecure code with access to GPT.
https://t.co/53IKR0QjZB
@Zane97070545@BrettFromDJ The capital required to start each is drastically different also. You can start a service company with a couple $100, but you need many $1,000’s for SAAS, and there is more risk.
@Zane97070545@BrettFromDJ True, but there’s no doubt that SAAS companies are harder to make profitable. People know they want services such as roof repairs, but not with SAAS.
It comes down to economies of scale. Service companies have very poor scaling, while SAAS have great, but start off bleeding.