LLMS - AI Dynamics

Precision Impact on Position Interpolation in ALiBi Embeddings

By

–

22 December 2023 22h11

TECHNICAL BLOG: We are excited to share some of our observations on the impact of precision on position interpolation and extrapolation techniques. We study this behavior on the ALiBi positional embedding first published by @OfirPress et al [1] and subsequently used in

→ View original post on X — @sambanovaai

22 December 2023

Apple’s AI Strategy: Building In-House Models Over GPT-3

By

@soumithchintala

–

22 December 2023 21h49

IMO apple will not call gpt3 because NIH and expense — and the local models (1B to 3B range) won't be as good. I think they'll build their own large model and local models and use them smartly (auto-routing or something). Maybe they'll fork Mistral's latest.

→ View original post on X — @soumithchintala

22 December 2023

5 Generative AI Use Cases Companies Can Implement Today

By

@miketamir

–

22 December 2023 21h01

5 Generative AI Use Cases Companies Can Implement Today https://
bit.ly/3RbKvof #AI #MachineLearning #DeepLearning #LLMs #DataScience

→ View original post on X — @miketamir

22 December 2023

NVIDIA TensorRT-LLM Integration Achieves State-of-the-Art Inference

By

@databricks

–

22 December 2023 20h54

For the last six months, we've been collaborating with @nvidia to integrate TensorRT-LLM with our inference service, achieving state-of-the-art inference performance. Read how we did it together and how you can benefit from our collab https://
bit.ly/3Hd7vyR

→ View original post on X — @databricks

22 December 2023

Google highlights 2023 AI research breakthroughs across multiple domains

By

@googleai

–

22 December 2023 19h50

2023 has been an exciting year for research across Google. Today @JeffDean
, @demishassabis
, and James Manyika highlight some of this year's exciting research across #GenerativeAI, language models, quantum, health, #ResponsibleAI, and more! Check it out at https://
goo.gle/3TBsyCp

→ View original post on X — @googleai

22 December 2023

December Newsletter: New Open-Source Models and NeurIPS Highlights

By

@cerebras

–

22 December 2023 18h54

Our December Newsletter is out! Learn more about the following:
-Our new open-source models and code, including gigaGPT, CrystalCoder, and BTLM-Chat
-Our amazing time at NeurIPS, which included presenting several papers at workshops, running the 1st Multilingual Workshop, and

→ View original post on X — @cerebras

22 December 2023

Alter3 Demonstrates Spontaneous Motion Capabilities Using GPT-4

By

@therundownai

–

22 December 2023 18h06

12. Alter3 is capable of 'spontaneous motion' using LLMs like GPT-4. pic.twitter.com/xwL8CRHxDj
— The Rundown AI (@TheRundownAI) 22 décembre 2023

12. Alter3 is capable of 'spontaneous motion' using LLMs like GPT-4.

→ View original post on X — @therundownai

22 December 2023

Automatic LLM Routing with notdiamond-0001 Model

By

@langchain

–

22 December 2023 18h00

Automatic LLM Routing with `notdiamond-0001` notdiamond-0001 automatically determines whether to send queries to GPT-3.5 or GPT-4, depending on which model is best-suited for your task Built by @tomas_hk
, they trained notdiamond-0001 on hundreds of thousands of data points

→ View original post on X — @langchain

22 December 2023

Anyscale Benchmark Flaws Raise Concerns About Research Quality

By

@soumithchintala

–

22 December 2023 17h52

it pains me to see a poorly constructed benchmark coming from a credible source. I hope @anyscalecompute fixes things, and also consults other stakeholders before publishing such benchmarks. If I didn't know Anyscale closely, i would have attributed bad faith.
Lets dive into

→ View original post on X — @soumithchintala

22 December 2023