EU's AI Act, Google's Gamer Companion
14 March 2024: EU AI Act Enforced, AI Gamer Companion, Startup Jobs Dip, Funding Surge, Amazon's Listing AI, Prostate Cancer Breakthrough, China's Rail Efficiency
Good Morning,
Latest in AI: As Europe draws the regulatory line in the digital sand, Google's DeepMind rolls out an AI buddy for the solo gamer's lonely hearts club. Meanwhile, the job market shuffles a headcount, and AI takes a scalpel to healthcare's old ways. Let’s dive in.
1-Minute Roundup
EU's landmark AI Act imposes strict regulations, targeting emotion-recognition bans and high-risk AI systems, with fines of up to 7% of global revenue.
Google DeepMind's SIMA aims to enhance solo gaming experiences with a versatile, instructable AI gamer companion.
Carta's report reveals a net decline in startup employment, highlighting AI's role in reshaping workforce demands.
Prescient AI secures $10M Series A for ad spend optimisation on its AI platform.
Tierra Biosciences raises $11M Series A to advance its AI-driven protein synthesis platform.
Tavus announces $18M Series A for digital replicas and text-to-video content.
CodaMetrix bags $40M Series B to improve medical coding and claims processing with AI.
Bear Robotics lands a $60M Series C, exploring new markets with LG Electronics.
Physical Intelligence $70M seed funding aims to develop versatile AI-powered robots.
Amazon's new AI feature simplifies product listing creation from external links.
AI research identifies two distinct prostate cancer types, promising personalised treatment avenues.
Perplexity AI integrates Yelp data to enrich local search experiences.
China's high-speed rail network benefits from AI for efficiency and maintenance.
Microsoft's upcoming Copilot for Security leverages AI for enhanced digital protection.
Adobe Firefly and Google Gemini AI navigate challenges in generating accurate, diverse imagery.
Headlines
EU Passes Landmark AI Act Setting Global Precedent
The European Parliament has approved the AI Act, imposing comprehensive regulations on AI systems within the EU market. The legislation sets a precedent with bans on specific AI uses and introduces stringent requirements for high-risk AI systems.
Scope and Fines: The AI Act applies to all AI products in the EU market and imposes fines. For breaches involving banned systems, maximum fines of up to €35 million or 7% of global annual revenue from the previous year are levied. Minimum fines of up to €7.5 million or 1% of the global annual revenue for inaccuracies or misleading information. Noncompliance fines can target AI providers, deployers, importers, distributors, and certified organisations.
Bans: In the EU, penalties apply for using AI systems that manipulate behaviour significantly, categorise individuals based on sensitive biometric data, score people based on social behaviour, utilise real-time biometric identification for law enforcement, assess criminal risk based solely on profiling, scrape facial images for databases, CCTV footage or infer emotions in workplaces and educational settings.
Provisions for General-Purpose AI: Providers must maintain technical documentation and publish training content summaries. Models with "systemic risk" need safety evaluations, incident notifications, risk mitigations, and cybersecurity measures.
Enforcement Timeline: Bans are enforceable later this year, with other obligations phasing between next year and 2027. Though EU-specific, the law is expected to influence global AI regulation and practices due to the EU's market size.
Deepfakes and High-Risk AI: Requires clear labelling of deepfakes and mandates risk assessments and high-quality data use for high-risk AI applications like immigration and critical infrastructure.
How might this shape the development and deployment of AI technologies worldwide, especially regarding safety and ethical standards?
Google DeepMind's SIMA: The AI Companion for Gamers
Google DeepMind's SIMA (Scalable, Instructable, Multiworld Agent) offers a new solution to loneliness in gaming by providing a companion AI that collaborates with players, emphasising participation over victory. Designed to understand and execute natural language commands, SIMA adapts to various game environments, from Goat Simulator to No Man’s Sky. It aims to make solo gaming sessions feel less isolated by adding a responsive, human-like teammate.
Versatile Training: SIMA was trained on a diverse portfolio of nine video games, including "No Man’s Sky" and "Teardown," to adapt to different virtual worlds and tasks.
Instructable Design: SIMA operates based on natural-language instructions and screen images, using keyboard and mouse actions to interact with games, mimicking human input methods.
Evaluation: The AI was tested on 600 basic skills, such as navigation and object interaction. In the future, it plans to tackle more strategic, multi-step tasks.
Generalisation Ability: SIMA outperformed specialised agents in games it was trained on and showed promising performance in games it had not seen.
How do you see such AI agents influencing the development of virtual assistants and their integration into various real-world applications?
Startup Employment Declines for First Time in Years, Carta Reports
For the first time in at least five years, the total net head count at startups has decreased, with layoffs and resignations surpassing new hires in 2023, according to Carta. The shift, influenced by funding challenges and AI adoption, marks a departure from the trend of growing startup teams, a move towards greater operational efficiency, and a more brutal job market for seekers.
Net Head Count Drops: In 2023, startups saw more departures (286,195) than hires (267,818), nearly half of the departures due to layoffs.
AI's Role: Adopting artificial intelligence enables startups to maintain or enhance productivity with fewer employees, reducing hiring.
Hiring Trends: The focus of new hires has shifted towards engineering, sales, and operations, with a decrease in managerial and customer support roles.
Compensation Changes: While salaries remained stable, equity compensation for new hires significantly reduced, reflecting broader market adjustments.
How do these changes impact the innovation ecosystem and job seekers' strategies in the tech sector?
AI Deals
Prescient AI secured $10M in Series A funding led by Headline and CEAS Investments to enhance its AI-driven platform for optimising ad spend and revenue for omnichannel brands.
Tierra Biosciences secured $11M in Series A funding to enhance its AI-driven platform for rapid custom protein synthesis, catering to pharmaceutical, industrial, and agricultural sectors.
Tavus announced a $18M Series A funding led by Scale Venture Partners and introduced a developer platform featuring the Phoenix model for creating digital replicas and text-to-video content.
CodaMetrix secured $40M in Series B funding, led by Transformation Capital, to enhance its AI-driven solutions for medical coding and claims data. The company aims to reduce healthcare costs and administrative burdens.
Bear Robotics has secured a $60M Series C funding round led exclusively by LG Electronics, marking a significant move into new markets like intelligent warehousing and supply chain automation.
Physical Intelligence secured $70M in seed funding to develop AI-powered robots applicable across various fields. The company aims to create universal models & algorithms for a wide range of robotics, merging language model techniques with machine control methods.
Others
Amazon's AI Magic: Paste a Link, Create a Listing
Amazon introduced an AI feature that creates product pages from external links, streamlining the process for sellers to list products on the platform.
This innovation is part of Amazon's broader adoption of AI. It aims to enhance sellers' efficiency with automated descriptions and images while maintaining strict guidelines on content ownership.
AI Uncovers New Subtypes of Prostate Cancer
AI research reveals two unique types of prostate cancer, offering new paths for personalised treatment.
The findings could lead to genetic tests for more accurate diagnosis and tailored patient care.
Perplexity Chatbot Enhances Local Searches with Yelp Integration
Perplexity AI chatbot now integrates Yelp data, offering users detailed local search results, including reviews and maps, for restaurants and cafes, aiming to enrich the user experience with reliable information.
The collaboration with Yelp marks a step towards providing chatbot users with current and comprehensive local insights by adding value through real-world data.
AI Revolutionizes China's High-Speed Rail Network
China employs artificial intelligence to oversee its vast 45,000-kilometer high-speed rail system, improving operational efficiency and predictive maintenance with an accuracy rate of 89%.
This AI-driven approach has led to an 80% reduction in minor track faults and ensures the network's expansion continues smoothly, connecting more cities while maintaining high safety standards.
Microsoft Unveils AI-Powered Copilot for Enhanced Security
Microsoft is set to launch Copilot for Security globally next month, introducing an AI assistant designed to bolster security measures across its product range.
Available under a flexible pay-as-you-go model, this new offering allows businesses to seamlessly integrate advanced AI capabilities into their security protocols, catering to a growing demand for intelligent security solutions.
Adobe Firefly and Google Gemini AI Face Similar Controversies
Adobe's Firefly AI tool falls into similar pitfalls as Google's Gemini, generating historically inaccurate images and raising concerns about AI's grasp on diversity versus historical accuracy.
Despite Adobe's cautious approach with legally sound training data, Firefly's outputs, mirroring Gemini's blunders, underline a broader challenge in generative AI regarding responsible and accurate content generation.
Love the AIML newsletter? Share it with your friends, family, and colleagues who want to stay on top of AI.