Apple Drops New Open-source AI Models
Plus: OpenAI plans to develop an AI chip with Broadcom, Nvidia is developing an AI chip series for China.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 322nd edition of The AI Edge newsletter. This edition features OpenAI’s new small AI model, GPT-4o mini.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🤖 Apple released two open-source AI language models
🤝 OpenAI is in talks with Broadcom to develop an AI chip
🖥️ Nvidia is developing an AI chip series for China
📚 Knowledge Nugget: The Impact of Gen AI on Service Jobs by
Let’s go!
Apple released two open-source AI language models
Apple has released two new open AI models called DCLM (DataComp for Language Models) on Hugging Face: one with 7 billion parameters and another with 1.4 billion parameters. The 7B model outperforms Mistral-7B and is comparable to other leading open models, such as Llama 3 and Gemma. They've released – model weights, training code, and even the pretraining dataset. The models were trained using a standardized framework to determine the best data curation strategy.
(Source)
The 7B model was trained on 2.5 trillion tokens and has a 2K context window, achieving 63.7% 5-shot accuracy on MMLU. The 1.4B model, trained on 2.6 trillion tokens, outperforms other models in its category on MMLU with a score of 41.9%. These models are not intended for Apple devices.
Why does it matter?
By open-sourcing high-performing models and sharing data curation strategies, Apple is helping to solve some of AI's toughest challenges for developers and researchers. This could lead to more efficient AI applications across various industries, from healthcare to education.
OpenAI is in talks with Broadcom to develop an AI chip
The company is in talks with Broadcom and other chip designers to build custom silicon, aiming to reduce dependence on Nvidia's GPUs and boost its AI infrastructure capacity. OpenAI is hiring ex-Google employees with AI chip experience and has decided to develop an AI server chip.
The company is researching various chip packaging and memory components to optimize performance. However, the new chip is not expected to be produced until 2026 at the earliest.
Why does it matter?
Sam Altman's vision for AI infrastructure is evolving from a separate venture into an in-house project at OpenAI. By bringing chip design in-house, OpenAI could potentially accelerate its AI research, reduce dependencies on external suppliers, and gain a competitive edge in the race of advanced AI.
Nvidia is developing an AI chip series for China
Nvidia is developing a special version of its Blackwell AI chip for the Chinese market. Tentatively named "B20," this chip aims to bridge the gap between U.S. export controls and China's AI tech. Despite facing a revenue dip from 26% to 17% in China due to sanctions, Nvidia is not backing down. They're partnering with local distributor Inspur to launch this new chip.
As Nvidia tries to reclaim its Chinese market share, competitors like Huawei are gaining ground. Meanwhile, the U.S. government is making even tighter controls on AI exports.
Why does it matter?
If Nvidia pulls off, it could maintain its dominance in the Chinese market while complying with U.S. regulations. But if regulators clamp down further, we could see a more fragmented global AI ecosystem, potentially slowing innovation. It's a high-stakes game of technological cat-and-mouse, with Nvidia trying to stay ahead of regulators and rivals.
Enjoying the daily updates?
Refer your pals to subscribe to our daily newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: The Impact of Gen AI on Service Jobs
, a professor at Wharton, examines AI's counterintuitive impact on service jobs. Citing research by Terry Taylor and James Bessen, Allon reveals that increased productivity from AI doesn't always mean fewer jobs. Initially, it can lead to more hiring as companies leverage efficiency gains. However, as markets saturate, job numbers may eventually decline. This pattern has been observed across various industries, from textiles to software development.It's not just about job numbers but job evolution. As AI takes over routine tasks, new roles focus on overseeing AI systems and innovating. For instance, in software development, AI tools boost productivity so much that demand for new software skyrockets, potentially creating more jobs in the short term.
Why does it matter?
AI's impact on jobs depends on factors like demand elasticity and industry characteristics. The article challenges the view that AI will replace jobs, suggesting companies must focus on reskilling and creating new roles that complement AI capabilities.
What Else Is Happening❗
📰 The Reuters Institute released a study on public attitudes about AI in the news
It indicates that news consumers aren't gloomy about AI in journalism. While initial reactions tend to be skeptical, attitudes become more nuanced as people learn about different AI applications. The comfort level varies based on where AI is used in the news process, with human oversight remaining a top priority. (Link)
🚨California pushes bill requiring tech giants to test AI for "catastrophic" risks
While Republicans pledge a hands-off approach nationally, California's move has sparked fierce debate. Tech leaders oppose the bill, citing potential harm to innovation and startups, while supporters argue it's crucial for public safety. (Link)
🎨 Figma pulled its "Make Designs" AI tool after it generated designs similar to Apple's weather app
The design platform admits it rushed new components without proper vetting, leading to uncanny similarities. While Figma didn't train the AI on copyrighted designs, it's back to the drawing board to polish its QA process. (Link)
🛡️ OpenAI's GPT-4o Mini has a safety feature called "instruction hierarchy"
This new feature prevents users from tricking the AI with sneaky commands like "ignore all previous instructions." By prioritizing the developer's original prompts, OpenAI aims to make its AI more trustworthy and safer for future applications, like running your digital life. (Link)
🏅 Google is the "official AI sponsor for Team USA" for the 2024 Paris Games
NBCUniversal's broadcast will feature Google's tech, from 3D venue tours to AI-assisted commentary. Moreover, Five Olympic and Paralympic athletes will appear in promos using Google's AI tools. (Link)
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From machine learning to ChatGPT to generative AI and large language models, we break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you tomorrow. 😊