The GIANT AI Report
The fast-paced AI landscape may seem overwhelming, as we are constantly inundated with new developments and breaking news. The GIANT AI Report is here to break it down into digestible bite-sized pieces that are currently relevant to parenting and education.
PBS KIDS Embraces AI Those awkward pauses you may know from shows such as Blue’s Clues and Dora the Explorer will soon become extremely laughable parts of our past (more so than they already are). PBS Kids is introducing AI-assisted characters that children can actually speak to and problem solve with in real time! The research and concepts that birthed this idea have been part of a long web of research stemming back to PBS’s creation. As “unnatural” as AI may be, this seems like a very natural next step for educational children’s television.
Gen Z Driving Early Adoption of Generative AI A study by Ofcom in the UK has found that teenagers and children are more likely to use generative artificial intelligence (AI) than adults. Approximately 79% of online teenagers aged 13-17 use generative AI, with 40% of children aged 7-12 also adopting the technology. Snapchat My AI is the most popular tool among children and teens, with 51% of online 7–17-year-olds using it, while ChatGPT is the most widely used generative AI service among internet users aged 16 and above (23%).
EU Passes Landmark AI Regulation Deal Late last month, the United States, Britain and more than a dozen other countries unveiled the first detailed international agreement aiming to create AI systems that are "secure by design." This agreement was followed by the EU's recent provisional agreement on AI regulations furthers its position as the first major global power to establish laws governing AI use. The accord mandates transparency for foundational models and general-purpose AI systems before entering the market, addressing issues such as real-time biometric surveillance and imposing fines for violations. While facing criticism from industry and privacy advocates, the legislation is viewed as a potential model for other governments navigating the delicate balance between reaping the benefits of AI and the necessity for regulatory oversight, highlighting the EU's pioneering role in shaping global AI governance.
AbSci and AstraZeneca Sign Deal on AI-Powered Cancer Drug AstraZeneca has entered a deal with AbSci Corporation, valued at up to $247 million, to develop an AI-driven antibody therapy for an undisclosed cancer. AbSci, a pioneer in healthcare-focused generative AI, utilizes its New York lab to analyze millions of protein samples, enabling cost-effective and time-saving drug trials, particularly in rare diseases. AstraZeneca's focus on cell therapies, including antibody therapies for cancer treatment, aligns with a growing trend in the pharmaceutical industry. Despite over $250 billion invested in AI-related deals in 2023, concerns over AI's efficacy in drug development and reported failures in the first wave of AI-designed drugs have contributed to a decrease in overall AI deal value since 2019. Oncology is currently the leading sector in AI Deals in Pharma.
How Does Google Gemini Measure Up? Google's unveiling of its AI model Gemini, aimed at competing with OpenAI's GPT-4, initially impressed with strong benchmarks and flashy demos. However, closer examination has revealed discrepancies between the video demo and details provided in Google's tech blog. AI developers found that Gemini's capabilities were less revolutionary than they appeared, and some even recreated the demo using GPT-4 Vision. Additionally, critiques have been raised about the benchmark data, with Google using different methodologies to measure performance, making it difficult to compare results accurately. Despite these challenges, Gemini's multimodal capabilities and performance improvements in certain scenarios have been acknowledged.
Meta’s AI Characters Are Now Live Across its US Apps As robots inch closer to human understanding and reasoning, will we soon be chatting with AI-powered programs more than we already are? Who needs Siri and Alexa when you can strike up a conversation with the likes of Paris Hilton, MrBeast, Kendall Jenner, Tom Brady, Charli D’Amelio, and Snoop Dog? Meta has fully rolled out its 28 AI characters, including those based on celebrities like Paris Hilton and Snoop Dog, for users to chat with across WhatsApp, Messenger, and Instagram in the U.S. The AI characters, which include features like long-term memory to remember past conversations, aim to create deeper connections and extended conversational capabilities. Meta's characters will also support search powered by Bing, enhancing user interaction and providing competition with other AI character apps.
Ray-Ban Meta Smart Glasses Can Identify Objects and Translate Languages Meta has announced the early access rollout of multimodal AI features for its Ray-Ban smart glasses, allowing users to interact with the AI assistant through the device's camera and microphones. The features include the AI assistant describing and suggesting items based on what it sees and hears through the glasses' camera and microphones. Users can ask for translations, captions for photos, and summaries, with the early access test limited to a small number of people in the US.
OpenAI and Alex Springer Strike Unprecedented Deal To Offer News in ChatGPT OpenAI and Axel Springer have entered an unprecedented deal allowing ChatGPT to summarize news stories from media brands such as Politico and Business Insider. Axel Springer will provide content from its media brands as training data for OpenAI's large language models like GPT-4. This collaboration comes amid growing concerns and legal actions by publishers and creators against AI companies for alleged copyright infringement, with notable authors suing OpenAI in September.
Does ChatGPT Go On “Winter Break”? Some ChatGPT users have observed the model becoming "lazy" in its responses since late November, refusing tasks or providing simplified answers. OpenAI has acknowledged the issue but is uncertain about the cause. Speculations, including the "winter break hypothesis," suggest the AI's behavior may be influenced by seasonal patterns, and users have experimented with feeding the model different dates to observe responses. The situation highlights the complexity and unpredictability of large language models, adding an element of uncertainty to their behaviors.
Researchers Create AI Tool For Fraudulent Wine Could AI put an end to wine fraud? Scientists at The University of Geneva in Switzerland have been giving AI its own personal wine tasting to discern which chemical blends correspond with specific chateauxs, and so far the accuracy is 99%!