GENERATIVE AI

–

21 November 2023 23h38

I love the idea of special effects loras.

→ View original post on X — @databricks

Building Custom LLM for Unity Catalog Documentation

By

@databricks

–

21 November 2023 23h33

2 engineers, 1 month, less than $1,000

That’s what it took to build a bespoke #LLM for AI-generated documentation in Unity Catalog. See how this hackathon project built on off-the-shelf SaaS-based LLMs became a well-loved UC feature ⬇️ https://t.co/t2WCBsbaWU pic.twitter.com/XciGnslWzx
— Databricks (@databricks) 21 novembre 2023

2 engineers, 1 month, less than $1,000 That’s what it took to build a bespoke #LLM for AI-generated documentation in Unity Catalog. See how this hackathon project built on off-the-shelf SaaS-based LLMs became a well-loved UC feature https://
dbricks.co/40TkqPd

→ View original post on X — @animaanandkumar

LeanDojo: Reducing LLM Hallucinations in Theorem Proving

By

@animaanandkumar

–

21 November 2023 23h28

For natural language, reducing hallucinations is a tradeoff between utility and creativity. Not so, for theorem proving and other applications that can be formally verified and certified. Check out our LeanDojo https://
leandojo.org using LLMs including GPT4 interface that has

→ View original post on X — @martyswant

OpenAI Releases Major Update with Advanced Voice Feature

By

@martyswant

–

21 November 2023 23h27

Amidst all the chaos, @OpenAI still managed to roll out another big update. FWIW: I've been testing the voice feature via ChatGPT Plus since it became available and it's pretty neat.

Quality degradation in AI models above 25 parameters

By

–

21 November 2023 23h16

Oh nice! I was going to run some experiments – I assume the quality gradually drops off above 25? How does it tend to degrade?

SVD-XT 25 Frames Benchmark on A40 GPU

By

–

21 November 2023 23h11

Here's the longer SVD-XT 25 frames version.
This took 159s to run on an A40. pic.twitter.com/7PjACUcsiC
— fofr (@fofrAI) 21 novembre 2023

Here's the longer SVD-XT 25 frames version.
This took 159s to run on an A40.

→ View original post on X — @nandodf

Defining AI Scaling: Data Quality Parameters and Evaluation

By

@nandodf

–

21 November 2023 23h06

I think this question could be expanded to clarify what is meant by scaling, e.g., for some this involves increasing data and its quality, evaluation, and ingesting as much data as fast as possible into a model for entropy reduction/prediction. Scaling more parameters only.

Stable Video Diffusion Text-to-Video Render Performance Analysis

By

–

21 November 2023 23h04

First text-to-video render of Stable Video Diffusion (SVD) from a Midjourney input image.

I'm impressed with

– coherent movement
– video quality
– accuracy with original image

Shame the explosions didn't, uh, explode. pic.twitter.com/qDnZAg0hzF
— fofr (@fofrAI) 21 novembre 2023

First text-to-video render of Stable Diffusion (SVD) from a Midjourney input image. I'm impressed with – coherent movement
– video quality
– accuracy with original image Shame the explosions didn't, uh, explode.

→ View original post on X — @martyswant

OpenAI Microsoft Face Copyright Infringement Lawsuit Generative AI

By

@martyswant

–

21 November 2023 23h00

OpenAI and Microsoft are facing yet another proposed class action lawsuit that alleges copyright infringement related to generative AI. The new complaint was filed today in U.S. District Court in New York's southern district on behalf of @THR reporter @jsancton
.