AI Act Advances: EU putting a leash on AI

PLUS: ‘AI-Powered’ VC Firm Vela Emerges From Stealth With $25M Under Management

Good morning! Today we’re covering the EU’s progress on banning predictive policing and facial recognition following the endorsement of the AI Act, Google's About this Image tool, which identifies real images and AI-generated fakes and Vela Partners, an AI-powered venture capital firm founded by former Google senior program manager Yiit Ihlamur.

Lets get started!

Top News & Why It Matters

| EU Nears Ban On Predictive Policing And Facial Recognition After AI Act Vote

The EU is moving closer to a potential ban on predictive policing and facial recognition in public spaces, as MEPs recently endorsed new rules on artificial intelligence in the proposed Artificial Intelligence Act. This groundbreaking regulation, set to become the world's first comprehensive AI law, must pass a European Parliament vote in June before implementation.

Civil liberties groups, such as Fair Trials, celebrate the decision as a positive move for human rights, as predictive policing systems can inadvertently perpetuate racial and societal inequalities. In contrast, lobby groups representing Microsoft, IBM, and others call for further clarification on the AI value chain, business models, and usage differences in the Act. thenextweb

Why does this matter?

Depending on the final outcome of the vote, privacy experts could come into higher demand as organizations increasingly emphasize AI governance, leaning on the expertise of their privacy teams to operationalize AI initiatives without breaching legislative boundaries.

| New Google Search Tool Will Distinguish Real Images From AI-Generated Phonies

Picture this: Google is launching a new tool called About this Image to help you determine if an image is real or fake. Soon to be unveiled in the US, this feature will be accessible through the three-dot menu on Google's search results, Google Lens, and eventually, the Chrome browser.

About this Image will provide details such as when and where the image first appeared, and where else it can be found online, including on social media or fact-checking websites. This information is aimed at helping users make an informed decision about the image's authenticity.

Google also plans to extend the tool's capabilities to work beyond its own search tools and apps. The company will ensure that every one of its AI-generated images contains a markup in the original file to provide context when viewed on other websites. zdnet

Why does this matter?

As image manipulation techniques become more sophisticated, it's crucial for tech giants like Google to step up and create tools that can help users identify fake content. Developing an ecosystem in which image authenticity is transparent and easily verifiable is becoming more important than ever.

| ‘AI-Powered’ VC Firm Vela Emerges From Stealth With $25M Under Management

Yiit Ihlamur, a former Google senior program manager, founded Vela Partners, an AI-powered and product-led venture capital firm. Vela aims to accelerate innovation by leveraging automation and artificial intelligence in the VC space. The firm currently has $25 million under management and holds 32 portfolio companies. Using predictive algorithms, Vela identifies trends, sources opportunities, and detects threats to its investments by analyzing data from websites, social networks, and paid datasets like Crunchbase.

However, the use of algorithms in the VC realm raises concerns about bias, as they often amplify existing biases in the data they're trained on. Ihlamur acknowledges this issue, but doesn't offer a clear solution. Vela's differentiator is its game-like terminal, which is designed to help entrepreneurs, limited partners, and other VCs use its services. The firm's algorithmic models are publicly available on its GitHub repo. techcrunch

Why does this matter?

As more VC firms explore AI-powered investing, addressing inherent biases in algorithms will be a crucial step in creating a more equitable investment landscape.

| Together With WriteWell

WriteWell is your fastest route to becoming a world-class communicator. It’s a custom AI trained on thousands of high-quality writing samples that goes beyond the capabilities of ChatGPT and other writing apps to make your writing concise.

Get 50 free credits when you sign up now using this link.

Mastering Artificial Intelligence

| Our Top Picks

✅ LagePro: AI tool for writing, grammar, translation, and conversation with virtual characters. (website)

✅ A service where you can generate AI avatars of your pets from just one uploaded photo. (website)

✅ Nureply: AI email service with personalized cold emails and automated campaigns. (website)

✅ Krecicki: AI sales call analysis for personalized closing technique improvement. (website)

✅ AI creates ad copy for real estate professionals, saving time. (website)

| Lunch & Learn

📝 Text: What is Mac Learning? (link)

🎧 Audio: Privacy in the age of AI by Practical AI: Machine Learning, Data Science (link)

📽️ Video: Make an AI sound like a YouTuber (LAB): Crash Course AI #8 (link)

What Else Is Happening

| More News

🗞️ 3 Questions: Jacob Andreas on large language models. (source)

🗞️ Success at the intersection of technology and finance. (source)

🗞️ IBM launches watsonx studio to make deploying generative AI easier. (source)

🗞️ OpenAI peeks into the “black box” of neural networks with new research. (source)

🗞️ Self-Driving Cars Are Being Put on a Data Diet. (source)

🗞️ Google Cloud is building its most powerful supercomputer to date. (source)

🗞️ Wendy's Installing Drive-Thru AI Chatbot That Takes Your Order. (source)

🗞️ Meta announces generative AI features for advertisers. (source)

🗞️ Spotify Has an AI Music Problem—but Bots Love It. (source)

🗞️ Study: AI models fail to reproduce human judgements about rule violations. (source)

🗞️ Argonne’s Polybot Joins the Self-Driving Lab Club. (source)

| Best of AI Social Media

💬 Google's Bard Outperforms ChatGPT with 8 New Free Features (Twitter)

💬 Google expected to focus on AI at I/O conference. (Twitter)

💬 AI Czar: Transforming business with artificial intelligence (Twitter)

💬 "Midjourney: A Wild Experience" (Twitter)

💬 8 AI Chrome Extensions Boost Productivity (Twitter)

💬 1+0.9 = 1.9 when GPT = 4. This is exactly why we need to specify which version of ChatGPT we used (Reddit)

💬 Notes from a teacher on AI detection (Reddit)

💬 RefuseToChatGPT (Reddit)

| Research Review

Accessible Instruction-Following Agent (Full Paper)

Kairui Zhou

∑ Summary: We introduce UVLN, a novel cross-lingual vision-language navigation framework, improving intractability and accessibility for non-English languages with limited resources.

!? Practical implications: Implementing the UVLN framework in robotics may enable more accessible and inclusive human-robot interactions, fostering broader adoption and the development of new applications across multiple language-speaking industries and societies.

Autonomous Gis: The Next-Generation Ai-Powered Gis (Full Paper)

Zhenlong Li, Huan Ning

∑ Summary: Researchers propose Autonomous GIS, an AI-powered system leveraging large language models for spatial problem-solving, showcasing its potential with a prototype called LLM-Geo.

!? Practical implications: The development of autonomous GIS may lead to increased efficiency and accessibility of spatial analysis for non-experts, potentially transforming industries like urban planning, environmental management, and resource exploration by simplifying complex spatial data interpretation.

Written by Isaac R. Ward, Casey Clifton, and Alex Brogan.

Send us your feedback at [email protected]. We’re ready to listen.

Join the conversation

or to participate.