The week's developments on AI, safety and ethics, explained | 19.04.24

If you’re enjoying this briefing, sign up here — and share it with a friend. Reading time: ~5 minutes

This week in AI: The UK has finally begun tiptoeing toward AI regulation, the DSIT drafting new legislation despite Sunak's previous 'no rush' stance. Meanwhile, as bigtech pumps the brakes on their push for immediate regulation, UK watchdog CMA raises ‘real concerns’ over their cosy AI alliance — and ensuing market dominance. Can the UK keep up without tripping over its own feet?

We cover these stories, plus teargas + paintball shooting AI home security, the startup helping you rent your data to AI, and how to develop an emotionally intelligent chatbot…

- Charlie and the Research and Intelligence Team

P.S Limited availability for our CogX Transatlantic Accelerator at $795 (RRP $23,621). If you are or know a UK startup, they can apply now for over $20k worth of support to attend, exhibit, and network at the CogX Festival in LA on 7th May here.

Share your expertise! Want to be a guest contributor in our next issue? drop us a line at: editors@cogx.live.

Ethics and Governance

🇬🇧 The UK is finally considering AI regulation, with the Department of Science, Innovation and Technology actively drafting legislation. Despite this push, Prime Minister Rishi Sunak maintains a cautious stance, asserting there is ‘no rush to regulate’ AI models and platforms.

⏸️ Tech leaders '180 on AI regulation, despite once advocating for immediate AI regulation the call now is to 'slow down.' This shift reflects growing concerns about the potential stifling of innovation and the practical challenges of reaching consensus on regulatory frameworks

🕵️‍♂️ UK competition watchdog has ‘real concerns’ about bigtech AI dominance, particularly with foundational models. Citing a complex web of partnerships among giants like Google, Apple, Microsoft, Meta, Amazon, and Nvidia, the CMA fears these alliances could exploit the market.

🔎 Meta investigates explicit AI-generated images: The Oversight Board is focusing on images involving public figures on Instagram and Facebook, highlighting gaps in Meta's content moderation. Findings will be published soon, with public feedback currently being collected.

📊 Want to rent your data to AI? Startup Vana is making it possible by creating the Reddit ‘Data Digital Autonomous Organization’, where users can pool their Reddit data and collectively licence it for AI training, potentially sharing in the profits.

AI Dilemmas

🚫 AI-generated child pornography is circulating; how do we regulate it? In California, DA Erik Nasarenko is pushing for legal changes after encountering unprosecutable cases due to current law gaps. The proposed Assembly Bill 1831 aims to make it illegal to possess, create or distribute such content.

🛡️Armed AI home security system — is it safe? Startup OZ-IT has launched PaintCam Eve, an AI camera that autonomously fires paintballs or teargas at threats using motion detection and facial recognition. However concerns arise about its potential to mistakenly target bystanders.

🚀Enjoying our content? We’re also on LinkedIn — Follow us to stay in the loop with the week's most impactful AI news and research!

Insights & Research

💬 OpenAI has updated ChatGPT to be less verbose and more conversational, with enhanced capabilities in maths, logical reasoning, and coding. This upgrade may blur the lines between AI and human-written content, heightening concerns that human writers could be mistaken for AI. 

😊 How to develop an emotionally intelligent chatbot: This research shows the promise of combining deep learning with NLP to enhance dialogue systems, enabling them to detect and respond to user emotions effectively, including signs of pain.

🔒 We need to embed privacy in all stages of AI research: This article emphasises the critical need for robust privacy measures to protect personal data and uphold ethical standards, ensuring research integrity and public trust.

🤔 Can AI really change our material world? A few months ago DeepMind claimed their AI had discovered over 2 million new crystalline materials, with a team from Berkeley ratifying the results. However, a recent swarm of critiques are questioning the novelty and utility of these findings.

In case you missed it

Boston Dynamics unveil their ‘All New Atlas’, it’s equally horrifying and incredible — just watch the video:

✍️ Enjoying this newsletter?  Subscribe to our free weekly briefings on Preparing for AI, Cinema & AI, The Race to Net Zero, AI & DeepTech and The Future of Work

🚀 CogX Transatlantic Accelerator Early Bird offer closes on 15th April: If you are or know a UK startup they can apply for over $20k worth of support to attend, exhibit and network at CogX Festival in LA on 7th May here