AI Weekly Digest Issue #32

Welcome back to AI Weekly Digest #32, your go-to resource for the most important artificial intelligence news and events from Week 47 of 2023.

As an AI enthusiast, staying informed about the rapid advancements in the field is crucial for both personal and professional growth. Our mission is to bring you a concise and comprehensive roundup of the latest breakthroughs, innovations, and discussions shaping the world of AI.

Nov 23, 2023

Inflection Announced Inflection-2: The World’s Second Most Capable LLM

Inflection, an AI startup focused on developing personal assistant technologies, recently announced the completion of Inflection-2 – their newest AI model. Inflection-2 demonstrates significantly improved capabilities compared to Inflection’s previous model, Inflection-1, showcasing better factual knowledge, language generation, reasoning abilities, and more.

The model was trained on over 5,000 GPUs for approximately 1 trillion parameters, placing its scale on par with top industry models like Google’s PaLM 2. However, benchmark testing showed Inflection-2 surpassing PaLM 2 on most standard AI benchmarks, especially in areas like question answering, common sense reasoning, and language understanding.

Inflection states Inflection-2 will soon power their personal assistant product called Pi. Combined with optimized inference implementation and upgraded H100 GPUs, Pi should see increased speeds and reduced serving costs compared to what was possible with Inflection-1.

From inflection

DeepMind Launches Tool to Identify AI-Generated Content

Google’s DeepMind has developed a new tool called SynthID that can watermark and identify AI-generated images and audio. The tool works by embedding an imperceptible digital watermark directly into AI-generated content during creation.

The watermarking technology converts images into spectrograms and audio into waveforms to embed the watermark so it’s undetectable to humans but can still be identified by SynthID later. This allows SynthID to scan content and detect the watermark to assess if it was created by an AI model.

SynthID is currently being used to watermark outputs from DeepMind’s newest AI music model Lyria and limited Imagen text-to-image models. The tool is robust to modifications of AI creations like compression, filters or noise and provides confidence levels to help determine if parts of content were AI-generated.

From DeepMind

Nov 22, 2023

Google’s Bard AI chatbot can now answer questions about YouTube videos

Google has upgraded its Bard chatbot’s ability to understand and answer questions about YouTube videos. Previously, Bard’s YouTube extension could only find videos on requested topics. Now Bard can provide specific answers about video content. For example, if you ask about a recipe video, Bard can tell you how many eggs are used. Or if you’re watching a travel video, you can ask where a location is and Bard will try to identify it.

This feature comes shortly after YouTube started testing AI tools to converse about video topics and summarize comment discussions. It also follows Bard being opened up to teens in most countries for inspiration and problem-solving. The upgrade allows richer conversations with Bard about YouTube videos to better tap into the knowledge and entertainment the platform offers. As one of the top AI assistants launches into more advanced video comprehension, it aims to enhance the user experience.

From techcrunch

Microsoft’s AI-Powered Copilot for Windows 10 Now Available for Testing

Microsoft has made an exciting move by backporting the AI-powered Windows Copilot feature from Windows 11 to Windows 10. This means that even if you’re still using Windows 10, you can now experience the benefits of this intelligent assistant. The feature is currently available in the Windows Insider Release Preview channel for Windows 10.

From theverge

Sam Altman’s Ouster at OpenAI: A Letter Unleashes a Storm

According to sources, this revelation, codenamed “Q*” (pronounced Q-Star), led to Altman’s abrupt departure.

The letter, penned by staff researchers, raised concerns about the potential impact of this AI breakthrough. Altman’s rapid pace of development and the perceived risks associated with it prompted the board to take decisive action. However, the manner in which Altman was removed—without prior notice to key stakeholders—has left many scratching their heads.

From cnbc

Tim Cook Calls for “Rules of the Road” to Guide AI Development

In a recent interview, Apple CEO Tim Cook expressed optimism about the potential of artificial intelligence to improve lives, citing future health applications as one example. However, he also emphasized the need to establish regulatory “rules of the road” to prevent misuse of AI.

Cook stated that while AI is already present across Apple products, the rapid advances in generative AI models like ChatGPT necessitate thoughtful governance. He reported that Apple is working with governments to shape policy, and advocated for regulations in the next 12-18 months. Though optimistic about government efforts, Cook said most are currently “a little behind the curve.”

The Apple chief also contrasted his company’s careful approach to AI with other tech firms, suggesting Apple considers potential downsides more deeply. His call for near-term regulation indicates tech industry support for reining in harmful applications of AI, even as companies rush to capitalize on its commercial possibilities. With public concern rising over fakes and misinformation, policymakers are likely to echo Cook’s call for establishing AI “rules of the road” soon.

From decrypt

OpenAI has reached an agreement in principle for Sam Altman to return to OpenAI as CEO

Nov 21, 2023

ChatGPT with voice is now available to all free users

Stability AI Releases Stable Video Diffusion: A Foundation Model for AI Video Generation

Stability AI has released Stable Video Diffusion, their first generative AI model capable of creating original video content. Currently available as a research preview, this state-of-the-art model allows text prompts to generate 14 or 25 frames of video at customizable frame rates. The code is available on GitHub and model weights can be accessed on Hugging Face.

Early benchmarking shows the model surpassing proprietary alternatives in quality and user preference. While focused on research applications for now, Stability AI envisions many downstream uses cases spanning sectors like advertising, education, and entertainment. The release follows their strategy with Stable Diffusion for images, building an ecosystem of specialized models around a reliable base foundation.

Signups are open for upcoming web demos showcasing text-to-video functionality. Stability AI emphasizes responsible research use only at this stage, soliciting user feedback to improve safety and quality before an eventual full release.

From StabilityAI

Accelerating Drug Discovery with AI: NVIDIA and Genentech Join Forces

Genentech and NVIDIA have announced a first-of-its-kind collaboration to optimize each company’s platform and accelerate drug discovery using generative AI. The goal is to understand complex molecular patterns to improve the success rate of research and development, and deliver new medicines to patients faster.

NVIDIA will work with Genentech to optimize proprietary algorithms on NVIDIA’s DGX Cloud AI supercomputing infrastructure. This includes leveraging NVIDIA’s BioNeMo platform to customize and scale drug discovery models. The initial focus is on enhancing Genentech’s “lab in a loop” framework that uses experimental data to inform computational models and improve future molecular designs.

By combining Genentech’s expertise in drug discovery and development with NVIDIA’s leadership in AI computing, the companies aim to streamline the lengthy and costly process of bringing new therapeutics to market. Insights gained can further advance NVIDIA’s healthcare offerings to meet biotech industry needs. Together, they hope to unlock scientific innovation and transform the way new medicines are invented.

From nvidia

Anthropic Launches Claude 2.1 with 200K Context, Higher Accuracy

Anthropic has released Claude 2.1, the latest version of its AI assistant. Key updates include doubling the context window to 200,000 tokens so Claude can process lengthy documents like codebases and novels. Claude 2.1 also reduces false claims by 2x compared to the previous version, boosting reliability for real-world applications. A new beta tool use feature allows Claude to integrate APIs and data sources, orchestrating workflows and searches to assist users. On the developer side, Anthropic introduced system prompts so users can customize Claude’s personality and structure. The Workbench product also streamlines prompt testing. With significant gains in accuracy and honesty alongside more customization and context, Claude 2.1 aims to power more robust enterprise use cases while prioritizing safety.

From Anthropic

AI System Mirrors Brain Complexity: Self-Organization Unveiled

In a groundbreaking study, an AI system has demonstrated an astonishing ability to self-organize, mimicking the intricate features found in the brains of complex organisms. The research, published in the European Review, sheds light on how artificial intelligence can spontaneously develop structures akin to those observed in living brains.

From techxplore

OpenAI Turmoil: Workers Rally for Altman’s Return Amid CEO Shake-Up

In a dramatic turn of events, hundreds of OpenAI employees have issued an ultimatum: reinstate former CEO Sam Altman, or they’ll walk out the door. The company, known for its groundbreaking AI innovations, now faces a crisis as its workforce demands change.

More than 700 of OpenAI’s roughly 770 employees signed a letter, urging the entire board to step down and calling for Altman’s reinstatement. Their message is clear: without Altman, they’re ready to follow him to Microsoft, where he has already found a new home.

From cbsnews

Amazon’s Bold Commitment: Free AI Skills Training for 2 Million People by 2025

Amazon Web Services (AWS) has unveiled its ambitious plan to equip 2 million individuals with essential AI skills by the year 2025. This initiative, aptly named “AI Ready,” aims to democratize access to cutting-edge technology and empower learners from diverse backgrounds.

From aboutamazon

Nov 20, 2023

Emmett Shear Takes the Reins: OpenAI’s New Interim CEO

Emmett Shear, the co-founder of Twitch, has stepped into the spotlight as the interim CEO of OpenAI. His appointment comes after the removal of former CEO Sam Altman, and it marks a significant shift for the artificial intelligence company. In a post on X (formerly Twitter), Shear expressed his acceptance of this “once-in-a-lifetime opportunity” to lead OpenAI.

Previously, Shear held the position of chief executive officer at Twitch, a role he occupied since the platform’s inception as Justin.tv. Under his leadership, Twitch became a powerhouse in the live video streaming world. Now, with his transition to OpenAI, Shear brings his expertise and vision to the forefront of AI research and development.

This leadership change at OpenAI has garnered attention, especially given the recent speculation surrounding ousted CEO Sam Altman’s potential return. With Shear at the helm, the company is poised for new directions and innovations. As the interim CEO, Shear’s background in technology and his experience in building successful platforms make him a compelling choice for steering OpenAI’s future endeavors.

From theverge

Harnessing AI and Science to Combat Malnutrition: UK Leader’s Urgent Call

British Prime Minister Rishi Sunak has issued a clarion call to the global community, urging us to wield the formidable tools of artificial intelligence (AI) and scientific innovation in our battle against malnutrition. His vision is ambitious: to transform the landscape of food security and eradicate hunger through cutting-edge technology.

Sunak unveiled a groundbreaking science initiative that aims to revolutionize crop development. The United Kingdom will spearhead efforts to create flood-tolerant rice, disease-resistant wheat, and other resilient crops. By harnessing AI algorithms and scientific expertise, these crops will withstand the challenges posed by climate change, ensuring stable food supplies for vulnerable populations.

Sunak emphasizes that this endeavor transcends national boundaries. The fight against malnutrition demands a united effort, with countries collaborating to share knowledge, resources, and breakthroughs. By leveraging AI and science, we can create a sustainable future where no one suffers from hunger.

From usnews

Spotify Teams Up with Google’s AI for Enhanced User Recommendations

Spotify has deepened its partnership with Google Cloud, aiming to revolutionize the listening experience for its users. The streaming audio giant, which has been working with Google Cloud since 2016, is now leveraging the tech giant’s powerful artificial intelligence (AI) tools to elevate its platform.

From fortune

Microsoft Hires Ex-OpenAI Leaders Altman and Brockman to Lead New AI Group

Microsoft has welcomed two prominent figures from the AI world into its fold. Sam Altman and Greg Brockman, co-founders of OpenAI, are set to spearhead a new era of artificial intelligence at the tech giant.

From Techcrunch

Nov 19, 2023

Elon Musk said that OpenAI should explain why it fired Altman, citing AI’s potential dangers.

OpenAI Board in Talks with Sam Altman for Potential CEO Return

Just over a day since his surprise firing that sent shockwaves through the tech industry, the OpenAI board is reportedly engaging in discussions with Sam Altman to potentially return as CEO of the company. Altman, a prominent figure in the AI community, had previously stepped down from his position but is now in talks to reclaim the helm.

The sudden ousting of Altman had raised eyebrows, but the latest developments indicate a potential course correction. The board’s attempt to hit “Ctrl-Z” and bring Altman back into the fold reflects the company’s desire for stability and continuity.

From theverge

Meta Disbands Responsible AI Team

Facebook-parent Meta has disbanded its Responsible Innovation Team (RIT), a group that played a central role in addressing potential societal harms arising from the company’s products. The decision, confirmed by a Meta spokesperson, has raised questions about the team’s effectiveness and the company’s commitment to ethical AI.

The RIT, comprised of engineers, ethicists, and other experts, was tasked with identifying and mitigating concerns related to Facebook and Instagram updates. Their mission was akin to that of a doctor—a first-line generalist—referring issues to specialists as needed. The team’s work extended beyond short-term considerations, emphasizing long-term impacts.

From CNBC

Every week, we’ll meticulously curate a selection of stories from top AI media outlets and distill them into a digestible format, ensuring you stay up-to-date with the latest developments without having to spend hours browsing the web. From groundbreaking research to real-world applications, ethical debates to policy implications, AI Weekly Digest will be your essential guide to navigating the ever-evolving landscape of artificial intelligence. Join us on this exciting journey as we explore the future of AI together, one week at a time.