The ChatGPT Observer Edition 33

Zeyad Sweidan

Senior Leader | Program & Project Management | Agile Coach | Relationship Focused | Systems & Technology | Business Intelligence | System Integration | Data Governance & Security

发布日期: 2024年10月20日

If you're short on time, here's your one-minute summary of this week's key AI news and advancements:

?? Top Story: Adobe MAX 2024 unveils 100+ new Creative Cloud features, including AI-powered tools for design, video, and audio. Highlights include Generative Fill in Photoshop, new 3D vector capabilities in Illustrator, and Quick Actions for easy edits in Lightroom.

?? Key Statistics & Insights: Generative AI startups account for 40% of all venture capital funding in cloud companies. Venture funding for cloud startups is projected to reach $79.2 billion in 2024. In 2023-24, $56 billion was invested globally in generative AI firms, with 80% going to U.S.-based companies.

?? ChatGPT & OpenAI News: OpenAI launches a dedicated ChatGPT app for Windows and introduces the experimental "Swarm" framework for AI agent collaboration. Former Palantir CISO Dane Stuckey joins OpenAI to lead security. OpenAI and Bain expand partnership to develop industry-specific AI tools for businesses.

?? Corporate AI Developments: Apple introduces new iPad Mini with A17 Pro chip featuring advanced Apple Intelligence. Apple reportedly developing smart glasses to rival Meta's Ray-Bans. Meta releases AI model that can evaluate other AI models' work. Google Cloud announces general availability of Vertex AI Search for Healthcare.

?? AI Innovations: Mistral releases new AI models optimised for laptops and phones. Perplexity introduces Internal Knowledge Search and Spaces for enhanced collaboration. HeyGen launches Interactive Avatar for Zoom meetings. Emteq Labs unveils emotion-tracking smart glasses.

?? AI in Business: Walmart CTO bets on AI for evolving shopping habits. Atlassian emphasizes AI time savings for focused effort, not reduced work hours. Nasdaq turns to AI to enhance risk assessment for banks and insurers. Noom introduces AI-powered body scan feature to track muscle mass during weight loss.

? Responsible AI: The New York Times sends cease-and-desist to Perplexity AI over content use. Penguin Random House bans use of its books for AI training. Singapore releases guidelines for AI security and bans deepfakes in elections. Elon Musk's X updates privacy policy to allow AI training on user data.

?? For more details, read the full newsletter!

Adobe MAX 2024: Empowering Creatives with 100+ New Tools

Adobe MAX is the leading creativity conference, showcasing the latest technologies, trends, and innovations. In 2024, Adobe unveiled over 100 new Creative Cloud features and early-stage innovations across photography, video, audio, and design.

Adobe MAX 2024 Highlights

Adobe introduced 100+ Creative Cloud features, enhancing tools like Photoshop, Illustrator, and Premiere Pro. For instance, Photoshop now includes the Generative Fill feature, which allows users to add or remove elements from images with simple prompts, and Illustrator has gained new 3D vector capabilities, making complex designs easier to create. Key updates include expanded Firefly AI tools, GenStudio's innovations, and improved Frame.io collaboration. Adobe also launched initiatives to boost AI literacy for 30 million learners.

Lightroom Enhancements

Lightroom now features Quick Actions for easy edits and Generative Remove, which automatically identifies and removes unwanted elements, simplifying the selection process. Performance improvements include 2x faster navigation and enhanced Nikon tethered capture. Expanded HDR support enriches editing.

Photoshop Innovations

Photoshop updates include Distraction Removal, powered by Adobe Firefly's latest model, which removes distractions such as wires and unwanted background elements. Generative Fill, Expand, and Background streamline edits, while object-based editing simplifies workflows on the web.

Generative AI for Marketing

Generative AI accelerates content creation, enhancing workflows and personalization. For example, Adobe Firefly can generate customized marketing messages or visuals based on user preferences, allowing brands to connect with their audience on a more personal level. Adobe Firefly allows brands to develop quality content rapidly, supporting real-time innovation and improved customer engagement. Adoption of GenAI is expected to rise.

Adobe MAX Sneaks

Adobe Sneaks are early-stage innovation projects that showcase experimental technologies. Adobe introduced nine Sneaks projects, including:

Project Perfect Blend: Seamlessly blend people or objects into images.
Project Clean Machine: Remove unwanted flashes or obstructions from photos and videos.
Project In Motion: Convert text into custom animations.
Project Turntable: Rotate 2D vector art in 3D without losing its essence.
Project Scenic: Create 3D scene layouts from 2D images.
Project Remix A Lot: Turn sketches into polished designs.
Project Super Sonic: Add sound effects using prompts.
Project Hi-Fi: Quickly create high-quality images from screen captures.

Generative AI startups

Generative AI startups account for 40% of all venture capital (VC) funding in cloud companies.
Venture funding for cloud startups is projected to reach $79.2 billion in 2024, with much of this driven by AI.
In 2023-24, $56 billion was invested globally in generative AI firms, with 80% going to U.S.-based companies.
OpenAI raised $18.9 billion, the highest amount for any U.S. generative AI startup, followed by Anthropic with $7.8 billion.

U.S. Defense and Homeland Security Spend $700 Million on AI Projects

The U.S. Department of Defense (DOD) and Department of Homeland Security (DHS) have awarded $700 million in AI-related contracts over the past two years, representing a 20% increase. The DOD awarded $670 million to 323 companies, with ECS receiving the largest contract at $174 million. The DHS awarded $22 million to AI projects, tripling its previous spending. Companies like Palantir and Scale AI are key contributors to these efforts.

Read more: Fortune

ChatGPT Arrives on Windows as a Dedicated App

OpenAI has launched a preview of its dedicated ChatGPT app for Windows, available to ChatGPT Plus, Team, Enterprise, and Edu users. This early version allows users to chat, upload files and photos, and generate images with DALL-E 3. While it works on most Windows 10 machines, it currently lacks features like voice support and some integrations with OpenAI’s GPT Store. The app can also run as a small companion window alongside other apps for multitasking. A more complete version is expected later this year.

Read more: TechCrunch

OpenAI Unveils 'Swarm' Framework, Sparking Debate on AI Automation

OpenAI’s new “Swarm” framework, introduced as an experimental tool, enables interconnected AI agents to collaborate and complete tasks autonomously. This technology has the potential to revolutionise business operations by automating complex processes across departments. Swarm allows developers to create multi-agent systems for various applications. However, concerns about job displacement, security risks, and ethical issues arise with the increasing automation. OpenAI emphasises that Swarm is experimental and not yet ready for commercial deployment.

OpenAI and Bain Expand Partnership to Develop AI Tools for Businesses

OpenAI and Bain are expanding their partnership to create industry-specific AI tools, focusing on sectors like retail and life sciences. This collaboration aims to go beyond the general use of ChatGPT, providing tailored solutions to drive business transformation. Bain’s 13,000 consultants now have access to ChatGPT Enterprise, and the partnership is developing tools like AI-based shelf planning for retailers and document automation for life sciences companies. While selling AI tools to businesses has proven challenging, OpenAI now has one million paying business clients and expects further growth.

Former Palantir CISO Dane Stuckey Joins OpenAI to Lead Security

Dane Stuckey, former Chief Information Security Officer (CISO) at Palantir, has joined OpenAI as its new CISO. Stuckey will work alongside OpenAI's head of security, Matt Knight, to enhance security and compliance for OpenAI's products. Stuckey's experience with government contracts at Palantir may help OpenAI strengthen its relationship with the U.S. Department of Defense. OpenAI has been ramping up its security efforts, including recent work with the Pentagon and a job listing to build secure AI infrastructure.

Read more: TechCrunch

HeyGen Interactive Avatar for Zoom Meetings

HeyGen's latest update introduces Interactive Avatars that can attend multiple Zoom meetings simultaneously, 24/7. These avatars not only resemble you but also think, talk, and make decisions based on provided knowledge. Equipped with Open AI Realtime Voice integration, they are ideal for online coaching, support, and sales calls, enabling intelligent conversations. This feature is perfect for handling repetitive meetings while freeing up time for more important tasks.

Mistral Releases New AI Models Optimized for Laptops and Phones

French AI startup Mistral has launched its first generative AI models, called "Les Ministraux," optimized for edge devices like laptops and phones. These models, Ministral 3B and Ministral 8B, are designed for local, privacy-first inference, catering to use cases such as on-device translation and internet-less assistants. Mistral claims that these models outperform competitors like Meta’s Llama and Google's Gemma. Ministral 8B is currently available for research, while both models are also accessible through Mistral's cloud platform.

Read more: TechCrunch

Perplexity Introducing Internal Knowledge Search and Spaces

Perplexity has launched two major updates: Internal Knowledge Search and Spaces. Internal Knowledge Search allows users to search both public web content and their own internal knowledge bases, providing a unified research experience. Spaces, an AI-powered collaboration hub, lets teams work together on projects, customizing their AI assistant with preferred models and instructions. These tools aim to enhance productivity by consolidating research efforts, while maintaining privacy and security for sensitive data.

Apple Introduces New iPad Mini with A17 Pro Chip

Apple unveiled a new iPad mini powered by the A17 Pro chip, featuring advanced Apple Intelligence for context-aware assistance. Apple Intelligence is deeply integrated into iPadOS 18, using on-device AI to understand personal context and provide helpful actions, from language generation to image creation. Many models run entirely on the device, ensuring privacy, while larger tasks can use Apple's Private Cloud Compute for additional processing power. This allows for an intuitive, intelligent experience that enhances daily productivity.

Apple Reportedly Developing Smart Glasses to Rival Meta's Ray-Bans

Apple is working on new smart glasses designed to appeal to a mass market, with a planned release in 2027. The glasses will leverage Apple's billion-dollar R&D investment from its Vision Pro headset, and will compete with Meta’s Ray-Ban AR spectacles, which currently start at $299. Additionally, Apple is exploring AirPods with cameras to incorporate visual intelligence features. Meta may maintain dominance in the AR space due to Apple's delayed timeline, as it continues to refine its own hardware and ecosystem.

Lenovo Unveils AI Buddy: A Smart Assistant with a Friendly Touch

At the 2024 Lenovo Innovation Technology Conference, Lenovo introduced the AI Buddy, a smart home assistant designed to interact like a friend. With a rotating circular display showing expressive animated eyes, it offers a more personal user experience. The device features USB ports, a headphone jack, and supports AI Now, a personal assistant developed in partnership with Meta. The AI Buddy can manage tasks, reminders, and daily activities while learning user preferences over time, focusing on privacy and security.

Meta Releases AI Model That Can Evaluate Other AI Models' Work

Meta has introduced a new batch of AI models, including the "Self-Taught Evaluator," which can evaluate other AI models' work without human input. This tool uses the "chain of thought" technique to break down complex problems into smaller steps, improving accuracy in fields like science, coding, and math. Trained entirely on AI-generated data, this model offers a pathway toward autonomous AI systems that can self-improve without human feedback, potentially reducing the need for expensive human involvement in AI development. Meta's release contrasts with companies like Google, which have not made similar tools publicly available.

NotebookLM Update: Customisable Audio Overviews and NotebookLM Business

Google's NotebookLM now lets users customise Audio Overviews, adjusting the AI's focus and expertise to fit specific needs. The feature allows for more guided, relevant deep dives, while also enabling background listening for better multitasking. Additionally, Google announced NotebookLM Business, an enhanced version available through Google Workspace for enterprises and universities. Organisations can apply for early access via the pilot program starting today.

Google Cloud Announces General Availability of Vertex AI Search for Healthcare

Google Cloud has launched its Vertex AI platform for healthcare, integrating new features that help professionals search and analyze health records more efficiently. Powered by Gemini and MedLM, Vertex AI leverages generative AI to assist clinicians with tasks like summarizing patient data and extracting insights from various sources. Additionally, the platform's Healthcare Data Engine (HDE) is now globally available, enabling better interoperability and access to comprehensive patient records. The move aims to reduce the heavy administrative burden on healthcare workers, allowing them to focus more on patient care.

Google Shopping Introduces AI-Powered Experience

Google Shopping has undergone a major AI-driven transformation, offering personalized shopping experiences. The new AI features, including AI-generated briefs and dynamic filters, help users find the best products quickly by showing the most relevant items and top considerations. The platform also includes features like virtual try-ons, personalized feeds, and price tracking tools. This revamped experience aims to make shopping easier, faster, and more tailored to individual needs.

Alibaba Claims Its AI Translation Tool Outperforms Google and ChatGPT

Alibaba's international arm has launched an updated version of its AI-powered translation tool, Marco MT, claiming it surpasses Google, DeepL, and ChatGPT in translation accuracy. The tool, based on Alibaba’s own Qwen large language model, supports 15 languages and uses contextual clues to enhance translations. Targeting merchants on Alibaba's platform, it aims to help businesses create more accurate product listings for global markets. Alibaba expects significant demand from Europe, the Americas, and emerging markets. The tool has already been used for over 100 million product listings.

Nvidia's Nemotron 70B: Leading AI Innovation in Large Language Models

Nvidia’s new Llama-3.1-Nemotron-70B model excels in aligning AI responses with human preferences, scoring 94.1% on the RewardBench leaderboard. With reinforcement learning from human feedback (RLHF), it delivers high efficiency and accuracy despite its smaller size. The model is ready for enterprise deployment, adaptable across cloud and data center infrastructures, positioning Nvidia as a strong competitor in the AI space alongside OpenAI and Google.

Cognizant Enhances Neuro AI Platform with Multi-Agent Capabilities

Cognizant has added multi-agent functionality to its Neuro AI platform, enabling businesses to ideate and prototype AI applications without coding. The platform, which now uses pre-configured agents like the Opportunity Finder, Scoping Agent, and Model Orchestrator, helps clients build tailored generative AI applications. It supports both open and closed AI models, making it adaptable for various enterprise needs. This development highlights Cognizant’s push in the competitive AI consulting space.

Amazon and Databricks Sign Five-Year AI Chip Deal

Amazon has partnered with Databricks, enabling Databricks to use Amazon's Trainium AI chips for AI model customization. Amazon claims its chips offer 40% cost savings compared to competitors like Nvidia. Databricks continues to expand its AI services, aiming to make AI faster and cheaper for businesses. The companies already have a strong collaboration through AWS, where Databricks' customers generated over $1 billion in revenue.

Amazon and Google Sign Nuclear Energy Deals Amid AI Power Demands Surge

Amazon and Google are investing in small modular reactors (SMRs) to meet the rising energy demands of AI technologies. Amazon is dedicating $500 million to nuclear energy projects for its data centers, collaborating with Energy Northwest and Dominion Energy. Google has signed a deal with Kairos Power for nuclear energy from SMRs, with reactors expected online by 2030. Both companies aim to reduce carbon emissions while supporting their AI advancements.

Emteq Labs Unveils Emotion-Tracking Smart Glasses

Emteq Labs introduced Sense glasses, an advanced wearable that tracks facial movements to assess emotional states, aimed at improving mental health and dietary management. Developed with insights from facial paralysis patients, the glasses use sensors to track expressions like smiling or chewing. They provide real-time feedback, including alerting users about overeating. Although consumer availability is pending, a development kit is expected in December. Privacy and ethical concerns remain, but Emteq sees potential in healthcare and personalized well-being applications.

Lenovo Showcases Smarter AI for All at Tech World 2024

At Tech World 2024, Lenovo revealed a range of AI-powered innovations focused on personalization, productivity, and data protection. Key highlights include the Lenovo AI Now local AI assistant, the ThinkPad X1 2-in-1 Gen 10 Aura Edition with hybrid work features, and Lenovo Learning Zone, an AI-powered educational platform. Lenovo also introduced exciting AI concepts like AI Buddy and enhanced AI security with ThinkShield?. These advancements aim to transform work, learning, and home life with smarter, more secure AI solutions.

领英推荐

?? AI Frontiers: Siri, OpenAI, and ChatGPT Upgrades

Generative AI 9 个月前

The AI chatbot innovation race is heating up

Freshworks 1 年前

Embracing AI with Microsoft Tools: A Transformative…

Professional Labs 1 年前

Human Native hires former Google execs to build AI data marketplace

London startup Human Native AI aims to become the go-to marketplace for AI data, helping companies find, license, and sell high-quality datasets. Co-founded by ex-Google engineers James Smith and Jack Galilee, the company hired former Google execs Madhav Chinnappa and Tim Palmer, alongside IP expert Matt Hervey. With a $3.6 million seed fund, Human Native seeks to standardize AI data licensing deals and help smaller publishers monetize their data. The company focuses on creating efficient licensing and governance tools for AI data transactions.

Read more: Fortune

AI Video Scraping: A New Frontier in Data Extraction

AI researcher Simon Willison introduced "video scraping," a technique where AI models like Google's Gemini extract structured data from video recordings. Willison successfully pulled pricing details from 12 emails by feeding a screen-recorded video to the AI, achieving accurate results at a very low cost. This development hints at broader potential for AI in automating tedious tasks, though it raises privacy concerns. AI-driven video input capabilities could revolutionize how users interact with data and screens.

Blumhouse Tests AI With Meta Partnership on Short Films

Blumhouse partnered with Meta to produce three short films using Meta's new AI tool, Meta Movie Gen, which generates video and audio. This pilot program aims to gather feedback from filmmakers like Casey Affleck and Aneesh Chaganty on Meta’s text-to-video AI capabilities. Meta plans to continue testing these tools with creatives into 2025 to ensure responsible use and effectiveness for storytelling. The collaboration highlights the potential of AI in filmmaking while still in development.

Noom Introduces AI-Powered Body Scan Feature to Track Muscle Mass During Weight Loss

Noom has launched a new AI-powered body scan feature, allowing users to create a 3D avatar and track their body composition using just a 10-second smartphone video. The feature, developed in partnership with Prism Labs, helps users monitor lean muscle mass, a key concern with popular weight loss drugs like semaglutide, which can cause muscle loss. Noom’s solution also includes a program encouraging increased protein intake and resistance training to retain muscle while losing fat, providing users with a comprehensive health report and visualizing their progress.

Harvard Scientists Develop AI Model with 96% Accuracy for Cancer Diagnosis

Harvard Medical School has unveiled CHIEF, an advanced AI model capable of diagnosing and predicting outcomes for various cancer types with an accuracy of up to 96%. Trained on millions of images, CHIEF can detect cancer cells, predict genetic profiles, and forecast patient survival, surpassing current AI systems. It works across 19 cancer types, providing insights into tumor behavior, treatment responses, and patient survival. The model has the potential to revolutionize cancer diagnosis and treatment by enhancing accuracy and efficiency in clinical settings.

Nasdaq Turns to AI to Enhance Risk Assessment for Banks and Insurers

Nasdaq is leveraging AI to improve risk assessment processes for banks and insurance companies. This AI-driven approach aims to enhance the identification of potential financial risks, allowing for better decision-making in these sectors. AI models are being used to evaluate risks more accurately and efficiently, helping organizations in the financial industry adapt to complex, evolving challenges. As the financial industry continues to embrace digital transformation, the adoption of AI technologies is becoming integral to managing risk and ensuring more secure operations.

Atlassian says AI time savings about ‘focused effort’, not working less

Atlassian’s head of AI, Sherif Mansour, emphasised that AI time savings should focus on enabling teams to invest time in more important tasks, rather than reducing working hours. Atlassian's Rovo AI tool, launched earlier this year, has been helping teams boost productivity by organizing and summarizing data across platforms like Jira, Confluence, and Slack. TBC Bank reported a 40% time savings and 25% efficiency gain. Nearly 75% of Rovo beta users found it increased work speed.

Walmart CTO Bets on AI for Evolving Shopping Habits

Walmart's Chief Technology Officer, Hari Vasudev, highlighted the shift in shopping behaviors, with consumers increasingly favoring personalized experiences over traditional online searches. Walmart is leveraging generative AI and predictive models to create a customized shopping experience. New features will include personalized recommendations based on past purchases and trends. Walmart’s AI tools aim to improve both the customer shopping experience and operational efficiency. A generative AI-powered customer support assistant was also launched, enhancing customer service interactions.

Read more: Fortune

Microsoft Expands AI Tools for Healthcare

Microsoft has announced new AI tools tailored for the healthcare sector, focusing on improving clinical decision-making, patient care, and workflow efficiency. The tools include automated clinical documentation, patient communication enhancements, and data analytics powered by Azure AI services. These tools aim to streamline operations for healthcare providers and improve patient outcomes. Microsoft's AI-driven approach integrates securely with existing healthcare systems, helping manage complex data while maintaining compliance with privacy standards.

AI in Healthcare: Arianna Huffington on the Transformative Potential

At Fortune's 2024 Most Powerful Women Summit, Arianna Huffington discussed how AI can transform healthcare by focusing on daily behaviours between doctor visits. With personalised health insights, AI could address areas like sleep, nutrition, and stress management, offering hyper-targeted recommendations. Huffington’s Thrive AI Health aims to empower users through small, impactful actions. Alongside Instacart’s CEO, Fidji Simo, she highlighted AI’s potential to tackle systemic health issues and improve overall well-being through data-driven, personalised care.

Read more: Fortune

Dutch Health Minister Advocates for AI Revolution in Healthcare

Dutch Health Minister Fleur Agema envisions AI playing a transformative role in healthcare, aiming to reduce administrative burdens and alleviate staffing shortages by 2030. Pilot projects, such as those at Elisabeth-TweeSteden and St. Antonius hospitals, are already exploring AI applications in emergency care and cancer detection. While AI shows promise, experts caution that implementation challenges like resource requirements and integration into workflows must be addressed.

NOAA Uses AI for Faster and More Accurate Hurricane Forecasts

The U.S. National Oceanic and Atmospheric Administration (NOAA) is leveraging AI, particularly neural networks, to improve hurricane prediction accuracy and speed. These models enhance forecasting by detecting intensity changes and translating alerts into multiple languages for diverse communities. Though NOAA mainly uses machine learning, it is researching generative AI techniques to develop future forecasting capabilities. AI is also helping NOAA combat misinformation during storm recovery efforts.

AI-powered Auslan Avatar Helps Deaf Train Travellers

An AI-powered Auslan avatar developed by researchers in Sydney helps Deaf train travellers by translating audio announcements into sign language. Deaf commuters often face challenges during travel due to reliance on audio updates. The avatar provides real-time translations for key updates, enhancing accessibility and safety. This technology is being developed to address the unique needs of the Deaf community, aiming for broader applications, including AI systems that can fully "understand" Auslan for everyday use.

The New York Times Sends Cease-and-Desist to Perplexity AI Over Content Use

The New York Times has issued a cease-and-desist letter to Perplexity AI, demanding the startup stop using its content without permission. The Times argues that Perplexity has benefited from the use of its journalism without proper licensing. Perplexity defends its actions by stating it doesn't scrape content for AI model training but indexes web pages to surface factual content, arguing that "no one owns the copyright over facts." Perplexity is set to respond by October 30th, with the company expressing interest in working with publishers, including the Times.

Penguin Random House Bans Use of Its Books for AI Training

Penguin Random House has added a clause to the copyright pages of new and reprinted books explicitly prohibiting their use for training AI systems. This move makes the publisher one of the first major companies to formally address AI in its copyright policy. The clause also aligns with European Union laws, reserving the text from data mining exceptions. While this stance serves as a warning, it doesn’t change copyright law, much like a website’s robots.txt file. Other publishers, like academic firms Wiley and Oxford University Press, have already made deals for AI training.

Singapore Releases Guidelines for AI Security and Bans Deepfakes in Elections

Singapore has introduced new guidelines to secure AI systems, covering five stages of the AI lifecycle to mitigate risks like data breaches and adversarial attacks. The guidelines emphasize a "secure by design" approach and were developed with AI and cybersecurity professionals. Alongside this, Singapore passed legislation banning the use of deepfakes in election advertising, making it illegal to manipulate content that misrepresents candidates in a way that could deceive the public. Fines and penalties apply for violations. Additionally, a cybersecurity label for medical devices was launched to ensure their security when connected to networks.

Elon Musk's X (formerly Twitter) Updates Privacy Policy to Allow AI Training on User Data

X has updated its Privacy Policy, allowing third-party companies to use user data to train AI models unless users opt out. Starting November 15, X will share user data with collaborators, potentially creating a new revenue stream by licensing its data to AI companies. While users can currently turn off data-sharing with xAI's Grok chatbot, the option to opt out of third-party AI training may be introduced when the policy becomes effective. X also added terms that penalize organizations for scraping large amounts of content.

Read more: TechCrunch

YouTube Introduces 'Captured with a Camera' Label for Authentic Videos

YouTube has launched a new "captured with a camera" label to indicate if a video was recorded with a real camera, without major alterations. This label relies on the C2PA standard, which ensures video authenticity by verifying the video's origin and confirming that its audio and visuals haven't been altered. To trigger this label, videos must be recorded and processed using tools supporting C2PA version 2.1 or higher. This new feature is part of YouTube's efforts to increase transparency, especially as AI-generated content becomes more common.

EU AI Act Checker Reveals Big Tech’s Compliance Shortfalls

A new AI Act compliance tool developed by LatticeFlow reveals that major AI models from companies like Meta, OpenAI, and Alibaba are falling short of EU regulations in areas such as cybersecurity and discriminatory output. The "LLM Checker" scores models across categories and shows where companies need improvement to meet the AI Act. OpenAI’s GPT-3.5 Turbo and Alibaba’s Qwen1.5 scored low in discrimination, while Google-backed Anthropic’s Claude 3 Opus performed the best.

IMD Launches AI Safety Clock to Monitor Uncontrolled AI Risk

IMD’s AI Safety Clock evaluates the risk posed by Uncontrolled Artificial General Intelligence (UAGI). Currently, the clock is set at 29 minutes to midnight, symbolising proximity to a critical tipping point. While UAGI has not caused major harm, rapid AI advancements combined with regulatory challenges necessitate vigilance. The clock’s risk is measured across three dimensions: Sophistication (AI reasoning), Autonomy (independence), and Execution(interaction with the physical world). Effective regulations can potentially slow down or reverse the clock’s progress.

Anthropic Updates Responsible Scaling Policy

Anthropic has revised its Responsible Scaling Policy to strengthen AI risk governance. Key updates include new capability thresholds, enhanced safeguards, and refined evaluation methodologies inspired by safety case techniques. The policy aims to address risks related to autonomous AI research and CBRN (chemical, biological, radiological, and nuclear) capabilities. The company emphasizes proportional safeguards via AI Safety Levels (ASL) that scale with risk. The updated policy reflects a year's worth of insights and aims to guide safe AI development.

EY’s Commitment to Responsible AI

EY has announced its updated Responsible AI Principles, designed to ensure the ethical and effective use of AI technology across all industries. These principles are aligned with key global frameworks, such as those from NIST, ISO, and the OECD, as well as the EU’s high-level expert group on AI.

The framework outlines 9 guiding principles to ensure ethical and safe AI development. These include:

Transparency: Clarity in AI's decision-making processes.
Security: Ensuring the safety of AI systems from threats.
Compliance: Adherence to legal and ethical standards.
Sustainability: Promoting long-term responsible AI use.
Data Protection: Safeguarding personal data.
Fairness: Avoiding bias in AI decisions.
Reliability: Ensuring AI systems are dependable.
Explainability: Making AI outcomes understandable.
Accountability: Clear responsibilities for AI’s outcomes.

These principles empower organisations to develop AI technologies that are trustworthy and benefit society while managing risks effectively.

Top AI Companies Struggle with Poor Risk Management, says French Non-Profit

A recent report by SaferAI, a French non-profit, reveals that leading AI companies, including Anthropic, OpenAI, and Meta, lack robust risk management practices. While Anthropic and OpenAI scored moderately on risk identification, Meta, Mistral, and xAI were rated poorly in risk analysis and mitigation. The report emphasizes the need for better risk management as AI technology advances. The EU's AI Office is addressing these risks, although concerns remain about its staffing and technical expertise.

Australian Government Releases Submissions on AI Guardrails

The Australian Government published over 300 submissions regarding mandatory AI guardrails in high-risk settings. The feedback includes views from organisations like Atlassian, Commonwealth Bank, and Westpac. These guardrails aim to manage AI risks, improve public trust, and provide clarity to businesses on AI usage. This input will shape AI regulation in Australia, focusing on safe and responsible AI development and deployment.

8 Useful AI Tools That Make Life and Work Easier

Eight powerful AI tools designed to enhance productivity and simplify daily tasks. These tools showcase the practical benefits of AI in both personal and professional environments.

Brain.fm: AI-generated music to enhance focus and relaxation.
Genei: Summarises long texts into concise bullet points.
Beautiful.ai: Helps create professional presentations quickly.
ChatGPT: Generates text, answers questions, and more.
LanguageTool: A powerful AI-based proofreading tool.
Jasper.ai: Creates creative texts in multiple languages.
Murf: Converts text to speech with customizable options.
Fireflies: Transcribes and analyses meeting conversations.

What's Next? The Future with Bill Gates.

??? We'd Love to Hear from You!

Did you find this edition insightful? Like, share your thoughts, and ask us anything.

Stay curious, and until next time, enjoy exploring the vast possibilities AI offers.

???? Subscribe on LinkedIn The ChatGPT Observer Newsletter

As we explore the captivating world of AI technologies, it's important to note that the content of this article is a product of human effort, enhanced by AI tools and additional resources. We have made sure to properly attribute and link all referenced articles to their original sources. For the visuals:

Images sourced from external sites are credited within the article.
Other images were generated using DALL-E, MidJourney and IdeoGram with enhancements made in Canva. These AI-generated and enhanced images are designed to visually represent each section's key themes.