Google's New Cheaper, Faster AI Models
Plus: Alibaba and Nvidia unite to create advanced autonomous cars, OpenAI introduces advanced voice mode, Meta Connect 2024 updates, and more.
Hello Engineering Leaders and AI Enthusiasts!
This newsletter brings you the latest AI updates in just 4 minutes! Dive in for a quick summary of everything important that happened in AI over the last week.
And a huge shoutout to our amazing readers. We appreciate youš
In todayās edition:
š Alibaba and Nvidia unite to create advanced autonomous cars
š£ļø OpenAI introduces Advanced Voice Mode with 5 new voices
š¤ Googleās new AI models: 50% cost, 2x performance
š Meta Connect 2024: Orion Glass, Quest 3S, Llama 3.2, and more
š„ OpenAIās CTO quits amid non-profit removal rumors
ā” Googleās new AI system designs chips faster than humans
š Knowledge Nugget: The Rapid Adoption of Generative AI by
Letās go!
Alibaba and Nvidia unite to create advanced autonomous carsĀ
Alibaba and Nvidia are collaborating on an AI initiative to advance autonomous driving technology for Chinese automakers. Alibaba Cloud is integrating its proprietary Qwen portfolio of LLMs into Nvidia's Drive AGX Orin platform, which major Chinese electric vehicle makers use.
With Qwenās advanced capabilities in handling complex inquiries and processing visual intelligence, the new autonomous driving solution will also offer intelligent recommendations, ranging from information about nearby landmarks to proactively suggesting car headlights be turned on during certain conditions.
Why does it matter?
For the first time, two AI powerhouses are collaborating on autonomous driving, providing a massive boost for the car industry's advancement. Nvidia has also decided to use Alibabaās Qwen LLMs, which is a win-win situation for the open-source community.Ā
OpenAI introduces Advanced Voice Mode with 5 new voices
OpenAI has finally rolled out its Advanced Voice Mode (AVM) to all ChatGPT Plus and Teams subscribers this week. The new voice feature has several improvements, including five new nature-inspired voices (Arbor, Maple, Sol, Spruce, and Vale), better accent understanding, and smoother, faster conversations.Ā
OpenAI had previously faced some legal challenges with one of the voice options, "Sky," which sounded too similar to actress Scarlett Johansson's voice. The company has since removed that voice option. However, the rollout has yet to be available in several regions, including the EU, the UK, and parts of Europe.Ā
Why does it matter?
At a time when OpenAIās CEO Sam Altman has been making headlines for AI Agents and Superintelligence, AVM feels relevant and is a natural progression toward that effort. As we increasingly interact with AI agents, AVM could make them more natural and human-like.Ā
Googleās new AI models: 50% cost, 2x performanceĀ
Google has announced significant updates to its Gemini AI models. The company is releasing two new production-ready models, Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002, which offer improved quality across various tasksāfor example, a 20% boost in math-related benchmarks.
Additionally, Google has reduced the prices for its Gemini 1.5 Pro model by over 50% on input and output tokens for prompts under 128K tokens. The company has also increased the rate limits for Gemini 1.5 Flash to 2,000 RPM and Gemini 1.5 Pro to 1,000 RPM. The new models boast 2x faster output and 3x lower latency than previous versions.
Why does it matter?
Google constantly pushes the boundary of affordability and accessibility, targeting developers building AI apps. These production-ready AI model upgrades are a move in that direction. It will help developers create faster, more intelligent, cheaper applications.
Meta Connect 2024: Orion Glass, Quest 3S, Llama 3.2, and moreĀ
Meta unveiled several significant new products and updates at their annual developer conference:Ā
Orion: Meta's "most advanced glasses the world has ever seen" - a prototype for their first consumer full holographic AR glasses.
Quest 3S: A new $299 wireless VR headset that offers mixed-reality features at a lower price point than the existing Quest 3.
Llama 3.2: This is Meta's first vision model with capabilities for both text and images, available in 11B and 90B parameter versions.Ā
Meta AI Voice: A new voice assistant for Messenger, Facebook, WhatsApp, and Instagram that can use celebrity AI voices like Judi Dench and Awkwafina.
Updates to Ray-Ban smart glasses: Real-time AI video processing and language translation.
AI-powered features for Facebook and Instagram: Automated video dubbing and AI-generated content recommendations.
Why does it matter?
With groundbreaking open-source models, the most advanced AR glasses, and AI Voice chat integrated across platforms, Meta has solidified its position as a leader in AI and mixed reality. Meta has 500 million monthly active AI users, showcasing its dominance in the industry.Ā
OpenAIās CTO quits amid non-profit removal rumorsĀ
Mira Murati, OpenAI's Chief Technology Officer (CTO), has announced her departure in a post on X. Hereās how OpenAI CEO Sam Altman responded:Ā
Murati's exit comes amid rumors that OpenAI is planning to restructure into a for-profit entity and give CEO Sam Altman equity ownership. The non-profit OpenAI will continue to exist but will no longer control the for-profit subsidiary entirely. This change could raise concerns about OpenAI's accountability for AI safety as it pursues more advanced AI systems.Ā
Why does it matter?
This is a massive shake-up for OpenAI's top leadership after it lost Andrej Karpathy, Ilya Sutskever, Greg Brockman, and chief research officer Bob McGrew. This raises concerns about OpenAI's functioning and whether it can fulfill its promise of being unbiased.Ā
Googleās new AI system designs chips faster than humans
Google DeepMind has created a new AI system called AlphaChip that can design computer chips much faster than humans. AlphaChip uses reinforcement learning to place components on chip layouts in hours, compared to human designers, who take weeks or months. The AI gets better at designing chips the more it practices, similar to how humans improve with experience.
AlphaChip has already been used to design the latest generations of Google's Tensor Processing Units (TPUs), making them faster and more efficient. Major chipmakers like MediaTek are also starting to use AlphaChip to accelerate their chip design process.
Why does it matter?
AlphaChip represents a pivotal moment in chip design using AI that could transform industries. Its self-reinforcing nature will enable much faster design cycles, allowing chips to be developed with better performance, power efficiency, and area utilization.Ā
Enjoying the latest AI updates?
Refer your pals to subscribe to our newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the āShareā 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 Rapid Adoption of Generative AI
In this post,
presents the first nationally representative survey on the usage of generative AI at work and home. They found that 39.4% of adults aged 18-64 use generative AI, with 32% using it at least once a week and 10.9% using it daily. Usage is higher at home than at work but still significant in the workplace, with 28% using it.The author is surprised by the rapid adoption of generative AI, which has been faster than the adoption of personal computers and the Internet in their early years. He suggests that generative AI is a "general purpose technology" used broadly across occupations and job tasks.Ā Ā
Why does it matter?
The survey contradicts many experts' skepticism about the usage of generative AI. While itās results can also be crucial for businesses to understand how customers leverage AI, for policymakers to craft informed regulations, and for workers to prepare for potential job market shifts.
What Else Is Happeningā
š¬James Cameron has joined the board of directors of Stability AI. Cameron's expertise in merging cutting-edge tech with storytelling will transform visual media through AI.
š¤Snapchat has partnered with Google to power the next-generation AI features in its chatbot, My AI. The chatbot will use Gemini AI to understand and process text, images, and videos.
āļøMicrosoft has launched a new "correction" feature in its Azure AI Studio that can automatically detect and rewrite inaccurate content generated by AI models.Ā
š„Hollywood stars and the actors' union are pushing California's governor to sign the AI safety bill as the tech and entertainment industries clash over the legislation.Ā Ā Ā
š«Meta has decided not to join the European Union's voluntary AI Pact immediately. Instead, it focuses on complying with the EU's upcoming AI Act, which will take effect in 2026.
šØThe FTC is cracking down on companies making deceptive claims about AI and using AI to mislead consumers by selling fake reviews and creating bogus business opportunities.Ā
šAMD has released its first small language model, AMD-135M, which is trained on AMD Instinct MI250 accelerators using 670B tokens.
āļøGoogle has updated Gmail's smart replies using Gemini AI to capture the email's intent better and provide more appropriate automated responses.
š¢Airtable has launched an enterprise-focused AI platform to help companies integrate and deploy AI capabilities more efficiently and effectively across their business workflows.Ā
š¤A software architect found a way to bypass OpenAI's guidelines and tricked the Advanced Voice Mode chatbot into singing with him in a duet of The Beatles' "Eleanor Rigby."Ā
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 next week! š