GENERATIVE AI

→ View original post on X — @hardmaru

Sakana AI’s cost-efficient model merging with limited GPU resources

By

@hardmaru

–

31 March 2024 6h55

“At the time, @SakanaAILabs only had 16 GPUs, which cost $30K per month. That’s pennies compared to the $100M+ it took to train advanced models like GPT-4… Model merging isn’t perfect though. @SakanaAILabs is one of only a few firms who have attempted to automate this process.”

GPT-4 Turbo Output Limitations Explained

By

–

31 March 2024 6h50

That's input though – the docs say output for gpt-4-turbo is limited to 4096

→ View original post on X — @inma_martinez, 2024-03-31 04:16 UTC

Concerns about OpenAI Voice Engine: Cybercrime and impersonation risks

By

@inma_martinez

–

31 March 2024 6h16

When is OpenAI to deliver a real beneficial use of this? Why would I want to clone my own voice? All I can think of is cybercrime and false impersonation. #stopopenai #stopvoiceengine @PaloAltoNtwks OpenAI (@OpenAI) We're sharing our learnings from a small-scale preview of Voice Engine, a model which uses text input and a single 15-second audio sample to generate natural-sounding speech that closely resembles the original speaker. openai.com/blog/navigating-t… — https://nitter.net/OpenAI/status/1773760852153299024#m

Cohere Embeddings Input Types for AI Applications

By

–

31 March 2024 6h10

For Cohere embeddings it's input_type="search_document", "search_query", "classification", "clustering"

→ View original post on X — @emollick

Treating AI Chatbots Like People for Better Results

By

@emollick

–

31 March 2024 5h53

Working with AI is already weird and is going to get weirder. Many forces are combining to make working with AI more like working with people, for better or for worse. In fact, you basically need to treat chatbots like people to get them to work well.

CLIP: Multimodal AI Bridging Text and Image Understanding

By

–

31 March 2024 5h43

I guess CLIP is another example of this, where the two modes are text and binary images

Voyage AI Input Types for Embeddings Documentation

By

–

31 March 2024 5h40

Thanks, yeah it looks like they call these "input types" https://
docs.voyageai.com/docs/embeddings

RAG Implementation: Embedding Passages and Queries Separately

By

–

31 March 2024 5h38

Being able to embed your content as "passage" but questions people ask about it as "query" is useful for implementing RAG – a user's question might not naturally embed to a similar location as content that answers that question, this trick helps fix that

→ View original post on X — @bilawalsidhu

3D Gaussian Splats Training Locally on Nvidia GPU

By

@bilawalsidhu

–

31 March 2024 5h09

It never gets old to see your 3D scans brought to life right in front of you, especially when you're doing the training locally.

This is a 3D Gaussian Splat with about 30,000 iterations trained on my Nvidia GPU using a software called Postshot.

With it you can train splats and… pic.twitter.com/89T4dLoto6
— Bilawal Sidhu (@bilawalsidhu) 31 mars 2024

It never gets old to see your 3D scans brought to life right in front of you, especially when you're doing the training locally. This is a 3D Gaussian Splat with about 30,000 iterations trained on my Nvidia GPU using a software called Postshot. With it you can train splats and