4 | AI Regulation: US vs EU

White House reveals AI executive order, Unveiling China's Place in the AI Spectrum, New AI-Powered Robocop

Author

Wilson CELY
October 27, 2023

This week, discover the protective measures the EU has embraced with its AI Regulation Act, and delve into the tech-forward initiatives brewing in Biden's cabinet for the U.S.

Explore how Robots are harnessing AI to navigate freely across major urban landscapes.

Plus, get a glimpse of the magic unleashed at the recent Adobe Max conference, where Adobe sprinkled a touch of innovation to fuel creativity like never before. 📖✨

Our Menu :)

WEEKLY DIGEST
AI Rulebook: White House reveals AI executive order Monday

October 26th, 2023. The big talk in tech town is all about the new rules for AI. The US and Europe are drawing their lines on how AI should play in our lives, each with their own style. Let’s dive into what’s cooking on both sides of the pond!

What’s New: 

The Biden administration is set to reveal a significant AI executive order just before a notable AI summit in Britain, attended by global leaders, Silicon Valley bigwigs, and civil groups.

This order emphasizes evaluating advanced AI models before federal use and simplifying immigration for skilled AI workers, boosting the U.S.'s tech edge.

Key departments like Defense and Energy are exploring AI integration to uplift national cybersecurity.

The upcoming "Safe, Secure, and Trustworthy Artificial Intelligence" event by the President on Monday will build on prior commitments by major companies like OpenAI and Google to promote responsible AI practices.

Why Does It Matter?

This action by the White House marks a transition towards stricter governance and oversight of the technology. Does this signify an end to the 'Wild West' era of AI development as swiftly as it emerged?

Similarities and differences with the EU approach

  1. Scope of Governance:

    • U.S.: Primarily targets federal agencies' deployment of AI, with a focus on the public sector.

    • EU: Encompasses a broader array of AI applications across both public and private sectors, indicating a wider scope of governance.

  2. Technological Advancement:

    • U.S.: Aims to ease immigration barriers for highly-skilled AI workers to bolster technological innovation.

    • EU: Lacks a similar provision, showing less emphasis on attracting global AI talent.

  3. Ethical and Consumer Protection Considerations:

    • U.S.: While not overtly tackling ethical issues, mandates federal agencies to adhere to existing privacy and discrimination laws.

    • EU: Places a strong emphasis on consumer safety, ethical considerations, data privacy, and algorithmic fairness.

  4. National Security and Cybersecurity:

    • U.S.: Prioritizes fortifying national cybersecurity through AI integration in various federal agencies.

    • EU: While not overtly focused on cybersecurity, includes provisions for AI applications in critical infrastructures and high-risk systems.

  5. Risk Assessment and Oversight:

    • U.S.: Employs 'red teaming' evaluations led by the National Institute of Standards and Technology (NIST) for risk assessments.

    • EU: Proposes a decentralized, multi-stakeholder oversight mechanism with a new AI board comprising representatives from all 27 member states.

  6. International Collaboration:

    • Both regions share a common objective to regulate and oversee AI, hinting at the potential for collaborative dialogue to develop more cohesive international standards for AI governance.

WEEKLY DIGEST
NVIDIA Unveils Eureka: LLM-Autonomous Reward Algorithm Generation for Robots

NVIDIA Research has introduced Eureka, an AI agent based on GPT-4, designed to autonomously create reward algorithms for training robotic systems. This innovative step was showcased by a robotic hand performing rapid pen-spinning tricks at a human-like level.

What’s New:

NVIDIA demonstrated Eureka training a simulated robot hand to spin a pen, open drawers, catch balls, and accomplish over 30 other varied tasks without any human involvement. This showcases Eureka's ability to create specialized algorithms and reward functions to guide the robot's learning through trial and error.

Why Does It Matter?

By automating a previously manual process, it increases task coverage and efficiency, outperforming human-authored rewards in 80% of tasks with a 50% average performance boost.

Hands-On Experience:

Key Details

  • Utilizes GPT-4 and reinforcement learning to autonomously create specialized robot training algorithms

  • Showcased on a simulated robotic hand mastering complex human-like manipulation

  • Automates reward function generation, previously a manual task in RL

  • Outperforms human-generated rewards in most tasks 

WEEKLY DIGEST
Unveiling China's Place in the AI Spectrum

Baidu Unveils Ernie 4.0 AI Model, Claims that it Rivals GPT-4

Ernie 4.0 is Baidu's latest AI marvel, poised to match the prowess of renowned models like GPT-4. It's a product of the escalating tech rivalry between the U.S. and China, extending the competitive frontier.

What’s New:
Ernie 4.0 demonstrated its potential in understanding complex verbal instructions, logic application, and content generation, set to enhance various Baidu services.

Why Does It Matter?
With claims of matching GPT-4's performance, Ernie 4.0 represents a substantial stride in global AI advancements, challenging the prevailing U.S. dominance.

Hands-On Experience:
Ernie 4.0 is yet to be publicly available, but its integration in Baidu's products like search and maps is eagerly anticipated.

Zhipu AI Secures $340M Funding Amidst Tech Race!

Zhipu AI, a blossoming entity from China's Tsinghua University, is making waves in the AI domain. Its recent accomplishment of amassing $340 million in funding showcases the robust financial backing it enjoys.

What’s New:
The funding will likely bolster Zhipu’s endeavors in AI, particularly in enhancing its bilingual conversational AI model, ChatGLM-6B.

Why Does It Matter?
The investment comes amidst a crucial time of tech rivalry, potentially giving Zhipu an edge in the AI competition.

Hands-On Experience:
Explore Zhipu's open-source models like ChatGLM-6B on their website.

ACTIONABLE TIP  
Dubai Police Unveil New AI-Powered Robocop

Dubai Police have showcased a new robocop aimed at boosting security and safety across the emirate. Unveiled at GITEX Global 2023, the robocop was designed and built by Dubai Police in collaboration with IBM.

What’s New:

The robocop features an array of high-tech capabilities powered by artificial intelligence. This includes facial recognition to identify wanted criminals, real-time access to criminal databases, and natural language processing to interact with the public.

Key Features

  • 360° cameras provide comprehensive monitoring of the surroundings

  • Touchscreen and microphone for two-way communication

  • Built on IBM Watson AI to analyze events and make decisions

  • Direct integration with police systems and control room

Deployment Plans

Dubai Police plan to deploy the robocops at busy areas like malls, tourist spots, and the Expo 2020 site. They will work alongside human officers to increase security presence across the emirate.

Wider Applications

If the robocop program is successful, Dubai Police aims to roll out more robots to conduct regular police work. This could include traffic monitoring, handling minor crimes, and expanding surveillance capabilities. 

ACTIONABLE TIP  
Ultra-realistic photos with Adobe Firefly

Adobe Firefly is a suite of generative AI tools developed by Adobe to create high-quality images, text effects, and color palettes from simple text prompts. 

What’s New:

The FireFly version 2 allows Ultra-Realistic-Image Generation using Text. 

We've put this to the test with the following prompt:

'Profile portrait of an Aztec indigenous woman, adorned in traditional attire, gracefully striding the runway at Paris Fashion Week.'

Look at the result 👀!

Why Does It Matter?

We can craft images that are nearly indistinguishable from genuine camera captures.
Look at the difference when compared to MidJourney or Dalle-3

Hands-On Experience:

It is completely free and easy to try out. Make sure to highlight this during your test

  • Visual Intensity slider to adjust the level of realism or surrealism of your images. For example, you can make a photo look more realistic by reducing the intensity, or more artistic by increasing the intensity More Info.

  • Generative Match, which allows you to create images that match the style of an existing image. You can either choose from a curated list of style images or upload your own image as a reference. More Info

ACTIONABLE TIP  
Prompt To Professional Vector Design

Adobe Illustrator has released a new AI-powered vector drawing feature that allows users to generate vector art by describing images with text prompts. This feature, called Text to Vector Graphic, is still in beta but has the potential to revolutionize the way designers create vector graphics. 

What’s New:

Vector graphics are scalable, resolution-independent images that are perfect for logos, icons, and other design elements. However, creating vector graphics can be time-consuming and challenging, especially for those who are not familiar with vector drawing software.

Why Does It Matter?

Ever dreamt of designing but found it too tricky? Vector graphics are a designer's best friend - they keep images sharp and clear, no matter the size. But making them can be a bit of a headache. Here's where Illustrator's new feature steps in, making vector drawing a piece of cake! Just type "a red apple with a green stem", and voila! You've got yourself a neat vector image of a red apple.

The Plus Side: 🤯 😱 🤯 

This isn't just good news for the pros. Even if you're new to design, now you can easily convert realistic images created by AI tools like Firefly and Dall-E 3 into vector graphics with Illustrator.

Hands-On Experience:

  • The Text to Vector Graphic feature is currently in beta and is only available to Adobe Creative Cloud subscribers.

  • The feature costs around 23 Eur/month, but there is a free trial available.

  • The feature is still under development and is not perfect, but it is very impressive and has the potential to be a valuable tool for designers.

TRAINING
Free Training with Stanford Professor Andrew Ng (Deep Learning AI) 🤖

Stanford professor and AI pioneer Andrew Ng recently announced an exciting new online course focused on generative AI. Titled "Generative AI: The Latest Breakthroughs and Ethical Considerations", this course aims to make cutting-edge generative AI concepts accessible to a broad audience.

Professor Ng, known for co-founding Google Brain and launching breakthrough AI courses on Coursera, will provide an introductory overview explaining how generative models like DALL-E 2, GPT-3 and AlphaFold work under the hood.

Link to learn more 👉

OUR SPONSOR
The Newsletter Platform Built for Growth

When starting a newsletter, there are plenty of choices. But there’s only one publishing tool built to help you grow your publications as quickly and sustainably as possible.

beehiiv was founded by some of the earliest employees of the Morning Brew, and they know what it takes to grow a newsletter from zero to millions.

The all-in-one publishing suite comes with built-in growth tools, customization, and best-in-class analytics that actually move the needle - all in an easy-to-use interface.

Not to mention—responsive audience polls, a custom referral program, SEO-optimized webpages, and so much more.

If you’ve considered starting a newsletter, there’s no better place to get started and no better time than now.

Try beehiiv absolutely free with no credit card required.

Thank you, see you next week!

Reply

or to participate.