a rocket is launching on a launch pad. a small bird sits contently amongst the billowing smoke and flames. The rocket is labeled "Twitter" in bold letters. --chaos 30 --ar 4:3 --style raw --personalize qko3v53 --v 6.1

Tech Papers, Training, and Development: Week Ending 07/05/2024

July 5, 2024

a rocket is launching on a launch pad. a small bird sits contently amongst the billowing smoke and flames. The rocket is labeled “Twitter” in bold letters. –chaos 30 –ar 4:3 –style raw –personalize qko3v53 –v 6.1

RouteLLM: An Open-Source Framework for Cost-Effective LLM Routing | LMSYS Org

https://lmsys.org/blog/2024-07-01-routellm

Mozilla Llamafile, Builders Projects Shine at AI Engineers World’s Fair – The New Stack

https://thenewstack.io/mozilla-llamafile-builders-projects-shine-at-ai-engineers-worlds-fair

Andrej Karpathy’s Keynote & Winner Pitches at UC Berkeley AI Hackathon 2024 Awards Ceremony – YouTube

AI scaling myths – by Arvind Narayanan and Sayash Kapoor

https://www.aisnakeoil.com/p/ai-scaling-myths

Training MoEs at Scale with PyTorch | PyTorch

https://pytorch.org/blog/training-moes

“100% Fully Software 2.0 computer. Just a single neural net and no classical software at all. Device inputs (audio video, touch etc) directly feed into a neural net, the outputs of it directly display as audio/video on speaker/screen, that’s it.” / X

100% Fully Software 2.0 computer. Just a single neural net and no classical software at all. Device inputs (audio video, touch etc) directly feed into a neural net, the outputs of it directly display as audio/video on speaker/screen, that’s it.
— Andrej Karpathy (@karpathy) June 30, 2024

Contra Acemoglu on AI – by Maxwell Tabarrok

https://www.maximum-progress.com/p/contra-acemoglu-on-ai

Deciphering Glyph :: Against Innovation Tokens

https://blog.glyph.im/2024/07/against-innovation-tokens.html

You’re Closer Than You Think: The Only 6 DNS Concepts You Really Need – JonahDevs

You’re Closer Than You Think: The Only 6 DNS Concepts You Really Need

“This is one of the coolest ideas for scaling synthetic data that I’ve come across. Proposes 1 billion diverse personas to facilitate the creation of diverse synthetic data for different scenarios. It’s easy to generate synthetic data but hard to scale up its diversity which is

This is one of the coolest ideas for scaling synthetic data that I've come across.

Proposes 1 billion diverse personas to facilitate the creation of diverse synthetic data for different scenarios.

It's easy to generate synthetic data but hard to scale up its diversity which is… pic.twitter.com/UR998d49hE
— elvis (@omarsar0) July 1, 2024

MInference: Million-Tokens Prompt Inference for LLMs

https://hqjiang.com/minference.html

“Qdrant engine v1.10 has been released with new powerful features. 🚀 ➡ 𝐔𝐧𝐢𝐯𝐞𝐫𝐬𝐚𝐥 𝐪𝐮𝐞𝐫𝐲 𝐀𝐏𝐈 with built-in Hybrid Search, a fusion merge of dense and sparse results, and multi-stage queries with re-scoring. ➡ 𝐌𝐮𝐥𝐭𝐢𝐯𝐞𝐜𝐭𝐨𝐫 𝐬𝐞𝐚𝐫𝐜𝐡 with late

Qdrant engine v1.10 has been released with new powerful features. 🚀

➡ 𝐔𝐧𝐢𝐯𝐞𝐫𝐬𝐚𝐥 𝐪𝐮𝐞𝐫𝐲 𝐀𝐏𝐈 with built-in Hybrid Search, a fusion merge of dense and sparse results, and multi-stage queries with re-scoring.

➡ 𝐌𝐮𝐥𝐭𝐢𝐯𝐞𝐜𝐭𝐨𝐫 𝐬𝐞𝐚𝐫𝐜𝐡 with late… pic.twitter.com/PfByCkniyn
— Qdrant (@qdrant_engine) July 2, 2024

RAM/projects/length_instruct at main · facebookresearch/RAM · GitHub

https://github.com/facebookresearch/RAM/tree/main/projects/length_instruct

Magic Insert: Style-Aware Drag-and-Drop

https://magicinsert.github.io

“Both nostalgic and uncanny valley at the same time

Both nostalgic and uncanny valley at the same time pic.twitter.com/qkbOhutJnO
— Peter Yang (@petergyang) July 4, 2024

“A recent Q* Paper – On MATH, surpasses GPT-4 and Gemini Ultra. 🔥 👨‍🔬 LLMs struggle with multi-step reasoning tasks due to errors accumulating across steps in auto-regressive generation. Existing solutions like prompting, fine-tuning, or using reward models have limitations. Q*

A recent Q* Paper – On MATH, surpasses GPT-4 and Gemini Ultra. 🔥

👨‍🔬 LLMs struggle with multi-step reasoning tasks due to errors accumulating across steps in auto-regressive generation. Existing solutions like prompting, fine-tuning, or using reward models have limitations.

Q*… pic.twitter.com/V9pNpTpBfv
— Rohan Paul (@rohanpaul_ai) June 30, 2024

Why AI Infrastructure Startups Are Insanely Hard to Build

https://nextword.substack.com/p/why-ai-infrastructure-startups-are

What’s the best interface for gen AI? It all depends on the use case | VentureBeat

What’s the best interface for gen AI? It all depends on the use case

“It’s officially been one week since Yi-Large launched on the Fireworks AI Playground! Have you had a chance to test the model out yet? If so, let us know what you like or dislike about Yi-Large in the comments 👇” / X

It’s officially been one week since Yi-Large launched on the Fireworks AI Playground! Have you had a chance to test the model out yet? If so, let us know what you like or dislike about Yi-Large in the comments 👇
— Yi-01.AI (@01AI_Yi) July 2, 2024

Simple Diffusion Language Models – YouTube

The Five Stages Of AI Grief – NOEMA

The Five Stages Of AI Grief

The shape of information – by Adam Kucharskihttps://kucharski.substack.com/p/the-shape-of-information

Heads up! You’ve scrolled to the end of this category. There may have been just one or two links (above), so go back up and double check to be sure you didn’t quickly scroll down past it.

Be Sure To Read This Week’s Main Post:

This week’s executive overview and top links are here:

AI News #40: Week Ending 07/05/2024 with Executive Summary and Top 65 Links

The post you just read is an deep dive extension of my weekly newsletter, This Week In AI, an executive summary of the top things to know in AI. Each week, I create an accessible overview for laypeople to feel confident they are conversant with the week’s AI developments. I include a curated list of must-click links of the week, to offer everyone a hands-on opportunity to explore the most intriguing updates in artificial intelligence across various categories, including robotics, imagery, video, AR/VR, science, ethics, and more. Beyond the overview, I post these topic-based deeper dives (below). If you haven’t read this week’s overview, I recommend starting there.

Credits/Sources

Most of these weekly links come from just a few prolific oversharing sources. Please follow them, as they work hard to find the news each week and they make it a lot easier for me to compile.