- Everyday AI
- Posts
- How Public AI Can Shape a Safer, Smarter Future for All
How Public AI Can Shape a Safer, Smarter Future for All
Google's Jarvis AI Leaks, Microsoft unveils multi-agent AI system, NVIDIA partners with Hugging Face for new AI robotic tools and more!
Everyday AI
November 06, 2024
š Subscribe Here | š£ Hire Us To Speak | š¤ Partner with Us | š¤ Grow with GenAI
Outsmart The Future
Today in Everyday AI
7 minute read
š Daily Podcast Episode: How can we make sure public AI can help shape our future? We have an expert from Mozilla break it down. Give it a listen.
šµļøāāļø Fresh Finds: Apple Intelligence image features get waitlisted, Google opens AI hub in Saudi Arabia and what Trumpās election means for AI. Read on for Fresh Finds.
š Byte Sized Daily AI News: Google's Jarvis AI Leaks, Microsoft unveils multi-agent AI system and NVIDIA partners with Hugging Face for new AI robotic tools. For that and more, read on for Byte Sized News.
š AI In 5: We found a secret hack that lets you talk to web pages using Microsoft Edge and Copilot. See it here
š§ Learn & Leveraging AI: Is it possible it create public AI that is safe and helpful for us all? We explain how it can be achieved. Keep reading for that!
ā©ļø Donāt miss out: Did you miss our last newsletter? We talked about the US Gov. using ChatGPT Enterprise, Apple allowing ChatGPT upgrade directly from iPhone and Perplexity's CEO offers AI replacement for striking NYT employees. Check it here!
How Public AI Can Shape a Safer, Smarter Future for All š
Ya know how in big cities thereās a mix of cars and buses?
Itās a symbiotic symphony of private vehicles and public transportation sharing the road.
Can AI be like that?
Nik Marda, the Technical Lead, AI Governance at Mozilla joins us to help us build that path.
Join the conversation and ask Jordan and Nik questions on AI governance here.
Also on the pod today:
ā¢ AI Integration in Daily Life š¤
ā¢ Trustworthiness and Concerns in AI Models š¤
ā¢ Funding and Investment Needs for Public AI š°
Itāll be worth your 32 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Upcoming Everyday AI Livestreams
Thursday, November 7th at 7:30 am CST ā¬ļø
Friday, November 8th at 7:30 am CST - How AI agents can bridge the gap to the future of enterprise work
Hereās our favorite AI finds from across the web:
New AI Tool Spotlight ā Video Ocean creates videos from text and images, Sona offers AI transcription notes and RivalSense lets you monitor any company with AI.
Apple ā Apple Intelligence image features have a waitlist.
Google ā Google is opening an AI hub in Saudi Arabia.
AI in Governance ā
With Trump being elected as U.S. President, hereās how it could affect the AI space.
The UK is launching a platform to boost businessesā trust in AI.
The White House is launching an R&D competition for AI in chip manufacturing.
Money in AI - Saudi Arabia plans $100B AI powerhouse to rival UAE.
Read This ā This new learning AI is teaching itself how to play a video game.
AI Models -
FLUX1.1 is adding high resolution capabilities.
Today, we are excited to share that we are adding high-resolution capabilities to FLUX1.1 [pro]. Extending its functionality to support 4x higher image resolutions (up to 4MP) while maintaining an impressive generation time of only 10 seconds per sample.
ā Black Forest Labs (@bfl_ml)
3:43 PM ā¢ Nov 6, 2024
Grok-beta is now on Poe.
Grok-beta is now on Poe!
An early preview of xAIās most intelligent model, Grok-beta offers cutting-edge reasoning capabilities and is well-suited for chat, coding, and complex multi-step tasks. (1/2)
ā Poe (@poe_platform)
7:02 PM ā¢ Nov 6, 2024
1. Google's Jarvis AI Leaks Ahead of December Launch š
Google has inadvertently revealed its upcoming AI agent, Jarvis, which is designed to assist users with web browsing in Chrome. The internal preview, briefly visible in the Chrome extension store, hinted at a tool that could manage tasks like online shopping and vacation bookings, positioning itself as a handy companion for holiday shopping.
While the prototype was quickly removed and remains unusable without special permissions, it underscores the growing trend of AI agents simplifying our digital lives.
2. Microsoft Unveils Magnetic-One: A New AI Rival in Task Automation š¤©
Microsoft has introduced Magnetic-One, a multi-agent AI system designed to tackle complex tasks traditionally handled by humans. Built on the AutoGen framework, this innovative system features an Orchestrator that coordinates specialized agents like WebSurfer and Coder to streamline operations ranging from web browsing to code execution.
As businesses increasingly seek efficient solutions, Magnetic-One promises to enhance productivity while addressing concerns raised by Salesforce CEO Marc Benioff about Microsoft's previous offerings.
3. NVIDIA Unveils AI Tools for Robotics and Hugging Face Partnership š¦¾
At the Conference for Robot Learning (CoRL) in Munich, NVIDIA has launched a suite of powerful AI and simulation tools designed to accelerate development for humanoid robots. The introduction of Isaac Lab's new workflows and advanced video data processing capabilities promises to streamline robot learning, making it easier for developers to create complex movements and interactions.
With collaborations like LeRobot between NVIDIA and Hugging Face, the focus is on enhancing open-source robotics research, potentially reshaping the landscape of AI development.
4. Perplexity Hits $9 Billion Valuation š¤
Perplexity, the AI search engine, is set to triple its valuation to an impressive $9 billion following a $500 million funding round, as reported by The Wall Street Journal on November 5. This meteoric rise from just $520 million at the start of the year reflects booming investor confidence in generative AI technologies.
However, the company faces legal challenges from news publishers alleging copyright infringement for using their content without compensation, raising questions about the ethics of AI content sourcing.
5. Meta AIās New Breakthrough in Video Generation š„
A new study from researchers at Meta AI and Stony Brook University has unveiled AdaCache, a transformative technique that enhances video generation speed while maintaining high quality. This innovative approach leverages adaptive caching and motion-based regularization to optimize the performance of video diffusion transformers, achieving processing speeds up to 4.7 times faster than previous methods.
With its plug-and-play design, AdaCache promises to revolutionize real-time video production, making it more accessible for developers and content creators alike.
Secret Microsoft Copilot Hack: Talk live with a web page
We found a secret hack to be able to talk to a website using the edge browser and Copilot to interact with the web in a new way!
We show you how it works and ways to use it.
Check out today's AI in 5.
š¦¾How You Can Leverage:
Ya know how in big cities thereās a mix of cars and buses?
Itās a symbiotic symphony of private vehicles and public transportation sharing the road.
Why canāt AI be like that?
Nik is the Technical Lead, AI Governance at Mozilla, the foundation behind Firefox thatās trying to increase accountability and trust on the internet and in AI models.
If youāre kinda new to the whole private/public model thang, hereās the sitch.
Right now, big enterprise companies and even small businesses are just using the private car equivalents of AI ā closed and proprietary large language models like ChatGPT, Google Gemini and Claude Anthropic.
But Nik makes the case that we need that private-public mix just like roads, or how thereās a PBS or NPR to the NBCs and CNNs.
So what the heck is Public AI?
How does it even work?
And why might it be a better option in some cases than Big Brand AI?
Glad you rhetorically asked. Hereās what ya need to know.
Letās dive in shorties.
Publicly. Let's dive into the silicon-laced tea leaves:
1 ā What the heck is Public AI? š¶āš«ļø
Public AI is already out and about.
You probably just havenāt used it.
Thereās Falcon 40B, an advanced ion engine developed by NASA's Evolutionary Xenon Thruster (NEXT) program. It's designed to improve propulsion efficiency and is expected to play a key role in future deep space missions.
Then thereās AlphaFold2, an AI-powered protein structure prediction tool developed by University of Copenhagen researchers affiliated with Google DeepMind.
Like public transportation, Public AI is usually affiliated with a government, nonprofit or foundation that wants to make more open, accessible and trustworthy AI systems for niche purposes.
While proprietary/closed models like OpenAIās, Anthropicās, Googleās and more are focused on benchmarks and profits, Public AI is focused on public good and accountability.
Try this:
Nik gave a simple breakdown of Public AI vs. Private AI.
He said Public AI hits these 3 criteria:
Public Good ā Itās open and accessible for the public good and resources
Public Orientation ā Itās centering on the needs of community and people (vs profit)
Public Use ā Itās prioritizing AI applications in the public interest
And generally, Public AI is either supported by governments, grants, nonprofits or research institutions.
But for now, most of the AI we use is likely closed source, proprietary models. Or, thereās Open Weights models like Metaās Llama that border on Open Source, but are still technically proprietary models that focus on profit.
Nik shared about another Public AI project from Mozilla called OpenVoice.
He said OpenVoice is a model where people can ādonateā their voices to make a public TTS model that is both more inclusive than closed models and more culturally accurate.
2 ā Public and Private AI can co-exist šÆ
Back to that whole traveling the road thang ā you donāt gotta like get a tattoo of a car and boycott the bus.
Both can exist. And both can serve their own purpose.
Nik said companies should actually try and use both. He said, AI strategy can be modular, or picking and choosing which models to use for which cases in larger enterprise scenarios.
As an example, a company might want to use GPT or Claude models for sales and marketing, but might want Public models for working with sensitive or secure on-prem databases.
In some instances, it can actually be faster, more secure and cheaper to find a mix.
Try this:
Need a deeper dive on the relationship and coexistence of Public AI and Private AI?
Samesies.
This article has a great breakdown of differences between closed and public AI models, and how companies can use both.
3 ā Donāt over-rely on the Big Guy š§
Letās be honest.
The Big Tech models are super cushy right now.
Theyāre ridiculously cheap and extremely accessible. And more than capable.
(Our hot take? Right now, Big Tech AI is racing for customers and are often burning cash in the process. Massive price hikes for future models are definitely gonna come. But for now, itās still the race for AI too cheap to meter.)
But Nik said they might not work out well for everyone in the future. Massive price hikes could change your companyās gameplan.
Or, a private LLM maker might completely change how their model operates under the hood, throwing off your companyās fine-tuned processes off track.
Try this:
What happens if youāre tapped into Claude, and they completely change prompt handling in their API? Or, if OpenAI has an outage for 24 hours.
AI model lock-in can be bad for your biz, especially if all aspects of your internal biz dev and customers/clients run through a single proprietary model.
What happens then?
Nik has a simple solution ā redundancy. He said Public AI models can work at the same time as close models. And if youāre already running some of your operations through a Public AI model, then if you need to switch over to that in the future, you shouldnāt hit any hiccups.
Wanna learn more?
The Mozilla Foundation just released an in-depth paper on Public AI thatās worth a read!
Now This ā¦
Let us know your thoughts!
Vote to see live results
Do you attend our livestreams?Every weekday, we bring you fresh AI insights, exclusive interviews, and breaking news with our Everyday AI livestream. |
Reply