🥟 Chao-Down #228 AMD unveils new MI300X chip, Google brings Gemini to Pixel phones, Meta releases new AI image generator trained on Instagram and Facebook, Apple open-sources new AI model framework

In the world of advanced AI semiconductors, very few companies can go head-to-head with NVIDIA. But with AMD’s new AI chip, the MI300X, the industry has a new and formidable contender.

According to AMD CEO Lisa Su, MI300X rivals NVIDIA’s H100 AI GPU in training LLMs but outperforms it by 1.4x when running Meta's Llama 2 for inference tasks.

We aren’t clear yet what pricing will be for the chips but Su said AMD's chip must offer a more competitive cost compared to NVIDIA’s ~$40,000 H100 to win over customers.

Some of those marquee customers include Meta, Oracle, OpenAI, and Microsoft as they seek to find alternatives to NVIDIA.

How will this industry shakeout? We’ll have to see as the AI wars continue.

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

Apple joins AI fray with release of model framework (The Verge)

Everything you need to know about Gemini, Google's new AI model (qz.com)

Meta’s new AI image generator was trained on 1.1 billion Instagram and Facebook photos (Ars Technica)

Gemini Nano now running on Pixel 8 Pro — the first smartphone with AI built in (google.com)

AMD releases new chips to power faster AI training (The Verge)

Why Spotify, Google, and OpenAI use the sparkles emoji for AI (qz.com)

Always be Learnin’ 📕 📖

Why Type Safety is Important (shuttle.rs)

Strategy, Not Technology, Is the Key to Winning with GenAI (Harvard Business Review)

Zone of Impossibility: Why Entrepreneurs Should Look For Impossible Challenges (Entrepreneur's Handbook)

Projects to Keep an Eye On 🛠

Mozilla-Ocho/llamafile: Distribute and run LLMs with a single file. (Github)

weaviate/Verba: Retrieval Augmented Generation (RAG) chatbot powered by Weaviate (Github)

adapter-hub/adapters: A Unified Library for Parameter-Efficient and Modular Transfer Learning (Github)

The Latest in AI Research 💡

TextDiffuser-2: Unleashing the Power of Language Models for Text Rendering (arxiv)

Language model self-teaching for domain adaptation (Morph)

Teach Llamas to Talk: Recent Progress in Instruction Tuning (gaotianyu.xyz)

The World Outside of AI 🌎

20 things we learned from the Epic v. Google trial (The Verge)

Wikipedia’s most visited page of 2023 is also its main threat (qz.com)

Nearly Half of Gen Z Say They're Anxious at Work Most of the Time: Survey (businessinsider.com)

Two paths after foster care - Two teens aged out of New Mexico’s child welfare system last year. What made their lives so different? (searchlightnm.org)

U.S. students' math scores plunge in global education assessment (Axios)

Streaming apps are trying to bundle their way out of customer disenchantment (Ars Technica)

One Last Bite 😋

Which jobs are most at risk of AI? New research from UK Department of Education!