a bustling Wall Street stock market trading floor. A stock ticker reads "Open Source" --chaos 30 --ar 4:3 --style raw --personalize qko3v53 --v 6.1

Open Source AI News: Week Ending 07/05/2024

July 5, 2024

a bustling Wall Street stock market trading floor. A stock ticker reads “Open Source” –chaos 30 –ar 4:3 –style raw –personalize qko3v53 –v 6.1

Gemma

“RT @armandjoulin: Gemma 2 27B is now the best open model while being 2.5x smaller than alternatives! This validates the work done by the te…” / X

Gemma 2 27B is now the best open model while being 2.5x smaller than alternatives! This validates the work done by the team and Gemini. This is just the beginning 💙♊️ https://t.co/qqgvt27bWI
— Armand Joulin (@armandjoulin) June 30, 2024

Meta/Llama

“Meta’s anticipated Llama-3-405B model is reportedly displayed in a recent WhatsApp update. The new version may be close to launching.

Meta’s anticipated Llama-3-405B model is reportedly displayed in a recent WhatsApp update.

The new version may be close to launching.https://t.co/5kvQKtPoD9
— Rowan Cheung (@rowancheung) July 1, 2024

“The only question is whether it will be open weights. If so, then we have the first open-ish GPT-4 class model. If it goes like other open model releases, there is going to be a burst of rapid innovation, much of it good, but some bad (expect quality phishing attacks at scale).” / X

The only question is whether it will be open weights.

If so, then we have the first open-ish GPT-4 class model. If it goes like other open model releases, there is going to be a burst of rapid innovation, much of it good, but some bad (expect quality phishing attacks at scale). https://t.co/RSNGwV18PT
— Ethan Mollick (@emollick) June 30, 2024

Meta drops AI bombshell: Multi-token prediction models now open for research | VentureBeat

Meta drops AI bombshell: Multi-token prediction models now open for research

Phi

“Great work on Phi-3! again by @Microsoft pushing some really cool performance improvement on Post Training updates to this family of models 🎉 Especially look at the massive JSON structured outputs performance improvement 1.9 → 60.1 🚀 The phi model family is a very exciting

Great work on Phi-3! again by @Microsoft pushing some really cool performance improvement on Post Training updates to this family of models 🎉

Especially look at the massive JSON structured outputs performance improvement 1.9 → 60.1 🚀

The phi model family is a very exciting… pic.twitter.com/vFuB5XNEoG
— Rohan Paul (@rohanpaul_ai) July 2, 2024

Qwen

“Qwen2 is the top model on the new @huggingface Open LLM Leaderboard. “Qwen2-72B-Instruct, one step above other models, scoring 43.02 on average (notably thanks to its performance in math, long-range reasoning, and knowledge). The current second-best model, Llama-3-70B-Instruct

Qwen2 is the top model on the new @huggingface Open LLM Leaderboard.

"Qwen2-72B-Instruct, one step above other models, scoring 43.02 on average (notably thanks to its performance in math, long-range reasoning, and knowledge).

The current second-best model, Llama-3-70B-Instruct… pic.twitter.com/OhYpDr6SLx
— Rohan Paul (@rohanpaul_ai) July 1, 2024

Other Open Source News

“🚀 We have released InternLM 2.5, the best model under 12B on the @huggingface Open LLM Leaderboard. 🔥 Outstanding reasoning capability 🚀 1M context window 🔧 Stronger tool use Try using it to build your amazing application!

🚀 We have released InternLM 2.5, the best model under 12B on the @huggingface Open LLM Leaderboard.

🔥 Outstanding reasoning capability
🚀 1M context window
🔧 Stronger tool use

Try using it to build your amazing application!https://t.co/VQ7oIMUcff pic.twitter.com/Nh8mbxWQR2
— InternLM (@intern_lm) July 3, 2024

“This is open source AI Anyone with a computer can do this in a couple of hours Next year anyone with a phone can do it in a few minutes The year after in a few seconds

This is open source AI
Anyone with a computer can do this in a couple of hours
Next year anyone with a phone can do it in a few minutes
The year after in a few secondspic.twitter.com/7HGmRyI4Yc
— Ate-a-Pi (@8teAPi) July 5, 2024

“@AndroidAuth Nomic AI launched GPT4All 3.0, the latest version of the leading open-source local LLM desktop app. The new version of the desktop app supports thousands of open-sourced models that you can run on all major operating systems, completely local and private.

Nomic AI launched GPT4All 3.0, the latest version of the leading open-source local LLM desktop app.

The new version of the desktop app supports thousands of open-sourced models that you can run on all major operating systems, completely local and private.pic.twitter.com/WCMTBfWAoC
— Rowan Cheung (@rowancheung) July 3, 2024

“Announcing GPT4All 3.0: The Open Source Local LLM Desktop App – Supports 1000’s of models and all major operating systems – Major UI/UX Improvements – Local File Chat with LocalDocs

Announcing GPT4All 3.0: The Open Source Local LLM Desktop App

– Supports 1000’s of models and all major operating systems
– Major UI/UX Improvements
– Local File Chat with LocalDocs pic.twitter.com/f0vMEUmWzL
— AndriyMulyar (@andriy_mulyar) July 2, 2024

“We are thrilled to announce Sentient’s $85M seed round, co-led by @foundersfund with @peterthiel, alongside @PanteraCapital and @hiFramework. This marks a pivotal step in aligning AI development toward a community-built open AGI platform. Our vision at Sentient is to ensure AI

We are thrilled to announce Sentient's $85M seed round, co-led by @foundersfund with @peterthiel, alongside @PanteraCapital and @hiFramework. This marks a pivotal step in aligning AI development toward a community-built open AGI platform.

Our vision at Sentient is to ensure AI… pic.twitter.com/6KyjjPcRUd
— Sentient (@sentient_agi) July 2, 2024

Unveiling of Moshi: the first voice-enabled AI openly accessible to all. – YouTube

“Kyutai Moshi – first real-time Audio LLM. Basically no delay – the LLM even interrupted the speaker a few times. It was actually a bit eager to answer very quick. 🙂 All to be open-sourced. Quality still a bit robotic though, but ok for v1. Pretty cool overall, congrats!

Kyutai Moshi – first real-time Audio LLM.

Basically no delay – the LLM even interrupted the speaker a few times. It was actually a bit eager to answer very quick. 🙂

All to be open-sourced. Quality still a bit robotic though, but ok for v1.

Pretty cool overall, congrats! https://t.co/JX0rLZT9Kv pic.twitter.com/wozmnr9zwX
— Lucas Beyer (bl16) (@giffmana) July 3, 2024

“RX1 Humanoid robot week 2 updates: @_buildspace @_nightsweekends – Opensource, low cost ($1,000) – Human-scale dual-arm manipulation – Teleoperation & machine learning – Pick and place applications for education/ research and Industrial/commercial POC

RX1 Humanoid robot week 2 updates: @_buildspace @_nightsweekends
– Opensource, low cost ($1,000)
– Human-scale dual-arm manipulation
– Teleoperation & machine learning
– Pick and place applications for education/ research and Industrial/commercial POC https://t.co/dZDJ7UYHyc pic.twitter.com/kXIlNFITQd
— Lethic (@lethic1) July 1, 2024

“Introduce Open-𝐓𝐞𝐥𝐞𝐕𝐢𝐬𝐢𝐨𝐧🤖: ⁣ We need an intuitive and remote teleoperation interface to collect more robot data. 𝐓𝐞𝐥𝐞𝐕𝐢𝐬𝐢𝐨𝐧 lets you immersively operate a robot even if you are 3000 miles away, like in the movie 𝘈𝘷𝘢𝘵𝘢𝘳. Open-sourced!

Introduce Open-𝐓𝐞𝐥𝐞𝐕𝐢𝐬𝐢𝐨𝐧🤖: ⁰⁣
We need an intuitive and remote teleoperation interface to collect more robot data.
𝐓𝐞𝐥𝐞𝐕𝐢𝐬𝐢𝐨𝐧 lets you immersively operate a robot even if you are 3000 miles away, like in the movie 𝘈𝘷𝘢𝘵𝘢𝘳.
Open-sourced! pic.twitter.com/629fQKtCWU
— Xuxin Cheng (@xuxin_cheng) July 2, 2024

Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs

https://cambrian-mllm.github.io

Text2Bricks: Fine-tuning Open-Sora in 1,000 GPU Hours | lego – Weights & Biases

https://wandb.ai/lambdalabs/lego/reports/Text2Bricks-Fine-tuning-Open-Sora-in-1-000-GPU-Hours–Vmlldzo4MDE3MTky

“The decreasing cost and increasing ability of open source AI image tools is leading to an explosion of interesting services and experiments. This is me playing with https://www.krea.ai/home  (free, for now?) which lets you “paint” in realtime with shapes.

The decreasing cost and increasing ability of open source AI image tools is leading to an explosion of interesting services and experiments.

This is me playing with https://t.co/dW3rswxg5e (free, for now?) which lets you "paint" in realtime with shapes. https://t.co/g9VgelRYjK pic.twitter.com/JggsBLAYvL
— Ethan Mollick (@emollick) July 5, 2024

Let’s Get Agentic: LangChain and LlamaIndex Talk AI Agents – The New Stack

https://thenewstack.io/lets-get-agentic-langchain-and-llamaindex-talk-ai-agents

“Reflection as a Service 🪞 A big issue for any agentic application is reliability – LLMs can error out and give problematic outputs. To get around this, developers oftentimes implement a reflection step which validates the output and self-corrects it if incorrect. This

Reflection as a Service 🪞

A big issue for any agentic application is reliability – LLMs can error out and give problematic outputs.

To get around this, developers oftentimes implement a reflection step which validates the output and self-corrects it if incorrect. This… pic.twitter.com/SdmXGcV7CK
— LlamaIndex 🦙 (@llama_index) July 4, 2024

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

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.