AI Weekly Digest Issue #31

Welcome back to AI Weekly Digest #31, your go-to resource for the most important artificial intelligence news and events from Week 46 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 17, 2023

Mira Murati Steps Up as OpenAI’s Interim CEO, Succeeding Sam Altman

OpenAI, a leading force in artificial intelligence, has announced a significant leadership change. Mira Murati, the company’s Chief Technology Officer, has been appointed as the interim CEO, replacing Sam Altman. Murati, who has been an integral part of OpenAI’s leadership team for five years, is recognized for her vital role in shaping the company into a global AI leader. Her expertise spans across the company’s research, product, and safety functions, making her a strong fit for her new role.

From OpenAI

Meta Introduced Emu Video and Emu Edit for Advanced Video and Image Editing

Meta has recently unveiled two significant advancements in generative AI: Emu Video and Emu Edit. These technologies mark a substantial leap in AI’s ability to augment human creativity and self-expression.

Emu Video introduces a streamlined method for high-quality text-to-video generation, leveraging diffusion models. This approach allows for efficient training and high-resolution video generation, favoring simplicity and effectiveness over prior methods requiring multiple models. It’s designed to generate detailed 512×512 videos at 16 frames per second, and has been preferred by users for its quality and adherence to text prompts.

Emu Edit, on the other hand, focuses on precise image editing. It enables free-form editing through text instructions, covering a wide range of tasks from local and global editing to color and geometry transformations. The technology aims for precision, altering only pixels relevant to the edit request, and has been trained on a dataset of 10 million synthesized samples. Emu Edit outperforms existing methods, offering unprecedented control and quality in image editing.

From Meta

Microsoft Releases AI Tool for Photorealistic Copying of Faces and Voices

Microsoft has unveiled an AI tool that promises to revolutionize the world of digital content creation. The tool, part of the Azure AI suite, enables the photorealistic copying of faces and voices with astonishing accuracy.

The technology, which combines advancements in deep learning and computer vision, allows users to generate highly realistic videos and audio clips. Imagine seamlessly swapping faces in a video or altering someone’s voice to match a different speaker—all while maintaining an uncanny level of authenticity.

From theguardian

Microsoft is bringing its AI Copilot to Windows 10 soon

In a move that promises to revolutionize the Windows experience, Microsoft is gearing up to introduce its AI Copilot to Windows 10. This powerful AI-driven assistant, already available on Windows 11, is set to make its debut on the older operating system, bringing a host of productivity-enhancing features.

From theverge

Nov 16, 2023

YouTube Partners with Artists to Explore AI Music Innovations

YouTube has been developing music-focused AI experiments in collaboration with Google DeepMind and a group of high-profile artists. Guided by a set of AI principles published earlier this year, these early experiments are designed to explore how AI can empower musical creativity, enhance artist opportunity, and deepen fan engagement on the YouTube platform.

The first experiment, Dream Track for YouTube Shorts, uses an AI model called Lyria to generate 30-second music tracks for Shorts based on prompts from creators. A small group of artists like Charlie Puth, Demi Lovato and T-Pain are participating to help shape this tool that matches creators with AI-generated music in the artists’ voices. Video clips demonstrate these artists testing out the Dream Track tool.

From Youtube

Forward Health Launches CarePods: An AI-Powered Doctor’s Office

Forward Health, a healthcare startup founded by Adrian Aoun, has launched CarePods, a self-contained, AI-powered doctor’s office.

The CarePods are equipped with advanced medical technology, including AI-powered diagnostic tools, and are staffed by licensed medical professionals. Patients can use the CarePods to receive a wide range of medical services, including primary care, urgent care, and specialty care.

The CarePods are also equipped with telemedicine capabilities, allowing patients to connect with medical professionals remotely. The CarePods are currently available in select locations, and Forward Health plans to expand the service to additional locations in the future

From techcrunch

Google’s Search Generative Experience (SGE) to revolutionize online shopping

Google announced its new Search Generative Experience (SGE) will help users find the perfect gift for their loved ones. The SGE will generate fake AI products that will help users find real gifts. The company aims to make online shopping more interesting and easier for users. The SGE will use machine learning algorithms to create a wide range of fake products that will be displayed in the search results. These products will be generated based on the user’s search query and will be accompanied by a description and a price. The SGE will also provide users with the option to purchase the real product that the fake product was based on. The company has not yet announced when the SGE will be available to the public, but it is expected to be released soon.

From theverge

Spotify to use Google’s AI to personalize podcast and audiobook recommendations

According to Reuters, Spotify has announced that it will be using Google’s AI technology to tailor its podcast and audiobook recommendations to individual users. This move is expected to improve the accuracy of Spotify’s recommendations and enhance the user experience. The AI technology will analyze the user’s listening history and preferences to suggest relevant content. This feature is expected to be rolled out in the coming months.

The article also mentions that Spotify has been investing heavily in its audiobook offerings, which were launched earlier this year. The company is optimistic about the potential of audiobooks and believes that it will be a great product that will open up more authors to more consumers.

From reuters

Nov 15, 2023

Google Photos Turns to AI for Smarter Photo Organization

In a recent update, Google Photos has harnessed the power of artificial intelligence (AI) to revolutionize how we manage our photo collections. The new features, rolling out on both Android and iOS, promise to make your photo organization experience more seamless and efficient.

Google Photos now employs AI algorithms to automatically tag and categorize your photos based on their content. Whether it’s recognizing objects, people, or specific details, the system does the heavy lifting for you. This means you can quickly find that sunset snapshot or locate all your beach vacation photos without manual effort.

From TechCrunch

Microsoft Unveils Custom AI Chip to Challenge Nvidia’s Dominance

In a strategic move to enhance its artificial intelligence (AI) capabilities, Microsoft has unveiled its custom-designed AI chip, codenamed Athena. This development aims to provide Azure cloud clients with more options for standard computing workloads and AI tasks, putting the tech giant in direct competition with market leader Nvidia.

From cnbc

Airbnb Acquires AI Startup for Just Under $200 Million

Airbnb, the global hospitality giant, has made its first acquisition since going public. The deal, valued at just under $200 million, marks Airbnb’s entry into the world of artificial intelligence.

The startup in question is GamePlanner.AI, a stealthy 12-person company led by industry veterans Adam Cheyer and Siamak Hodjat. While the specifics of GamePlanner.AI’s technology remain undisclosed, the acquisition signals Airbnb’s strategic intent to enhance its services through AI-driven innovations.

This acquisition aligns with Airbnb’s mission to revolutionize travel experiences. By integrating AI capabilities, the company aims to personalize recommendations, streamline bookings, and improve overall guest satisfaction. GamePlanner.AI’s expertise could play a pivotal role in achieving these goals.

From cnbc

Be My Eyes AI Offers GPT-4-Powered Support for Blind Microsoft Customers

Microsoft has teamed up with Be My Eyes, an app designed for blind and low-vision individuals, to enhance customer service accessibility. The partnership aims to empower blind users by providing them with direct technical assistance through the app.

Be My Eyes is a smartphone app that connects blind or low-vision people with sighted volunteers via live video calls. The app assists users in various tasks, such as cooking, locating lost items, reading letters, and even navigating public transportation. With over 900,000 users across 150 countries, Be My Eyes has become a vital resource for the visually impaired community.

From theverge

OpenAI is pausing new ChatGPT Plus sign-ups for a bit

Nov 14, 2023

Google AI Introduces Mirasol3B for Advanced Multimodal Learning

Google AI has unveiled Mirasol3B, a groundbreaking multimodal autoregressive model designed to learn from audio, video, and text data. This model addresses the complex challenge of processing inputs from heterogeneous modalities, which often differ significantly in synchronization and data volume. Traditional multimodal models struggle to integrate these varied inputs effectively, particularly for longer videos, due to the disproportionate volume of audio and video data compared to text.

Mirasol3B innovates by decoupling multimodal modeling into distinct autoregressive models, each tailored to the specific characteristics of the modalities it handles. The architecture comprises an autoregressive component for time-synchronized modalities (audio and video) and another for sequential but not necessarily time-aligned modalities (like text). This separation allows for more efficient and specialized processing of each data type. For instance, time-aligned modalities like audio and video are partitioned temporally to facilitate joint feature learning.

From GoogleAI

AI Churns Out Lightning-Fast Forecasts as Good as the Weather Agencies’

In the high-stakes world of weather forecasting, where precision can mean the difference between disaster and preparedness, artificial intelligence (AI) is emerging as a game-changer. A recent article published on Science.org explores how AI is revolutionizing weather prediction, rivaling traditional weather agencies in accuracy and speed.

From science

GPT-5 Is Officially on the OpenAI Roadmap Despite Prior Hesitation

OpenAI has officially placed GPT-5 on its development roadmap, signaling a bold step forward despite prior reservations. As the successor to the wildly successful ChatGPT, GPT-5 aims to be a major leap beyond its predecessor, GPT-4.

While initial reports hinted at a potential launch in 2024, the latest information suggests that GPT-5’s training has not yet begun. OpenAI had previously announced an “intermediate version” called GPT-4.5, slated for release in September or October 2023. However, the timeline for GPT-5 remains uncertain.

Artificial General Intelligence (AGI) is the holy grail of AI, and early rumors suggest that GPT-5 might inch closer to this elusive goal. Some experts speculate that it could achieve a level of intelligence akin to AGI, but the debate continues. OpenAI’s co-founder, Sam Altman, has confirmed that GPT-5 is not yet in training, emphasizing the challenges ahead.

From decrypt

YouTube Takes a Stand Against AI Musicians

YouTube is gearing up to tackle a growing phenomenon: AI-generated clones of musicians. The video-sharing giant, owned by Google, is taking steps to protect artists’ rights and preserve the integrity of music creation.

Artificial intelligence (AI) has made remarkable strides in music composition. However, some AI models have become so advanced that they can replicate melodies and even mimic musicians’ voices with astonishing accuracy. This raises concerns for the music industry, particularly for labels like Universal Music Group (UMG), which represents superstars like Sting, The Weeknd, Nicki Minaj, and Ariana Grande.

From theverge

AI Unveils Crystal Ball: Predicting Heart Attacks a Decade Ahead

Researchers from the University of Oxford have harnessed the power of artificial intelligence (AI) to peer into the future—up to a decade ahead. Their focus? Heart attack risk. The implications are profound: imagine being able to identify individuals at risk of cardiac events years before symptoms manifest.

The secret lies in cardiac CT scans. These high-resolution images of the heart provide a wealth of information, but their analysis has been limited by human capacity. Enter AI. By meticulously examining these scans, the algorithm can predict an individual’s risk of a heart attack or stroke over the next ten years. This isn’t science fiction; it’s science fact.

From theguardian

Nov 13, 2023

New Nvidia Supercomputing Chips to Fuel the Acceleration of AI

NVIDIA has unveiled its latest AI system, the HGX H200, powered by the H200 GPU. This announcement comes during the company’s Supercomputing 2023 event, where they showcased upgraded chips and a supercomputer designed to accelerate scientific research.

Ian Buck, NVIDIA’s Vice President of high-performance computing and hyperscale data center business, emphasized the critical role of computing in modern civilization. He pointed out that while CPU performance has grown significantly over the past two decades, incremental gains have led to disproportionate costs and increased power consumption. The new computing law, as framed by Buck, necessitates innovative solutions.

The NVIDIA H200 represents a leap forward. Not only does it deliver improved performance, but it also enhances cost and energy efficiency. With HBM3e memory, this GPU fuels the acceleration of generative AI and large language models. Its 14GB of memory processes an impressive 4.8 terabytes per second—nearly double the capacity and 2.4 times more bandwidth compared to its predecessor.

But that’s not all. The Grace Hopper superchips, a fusion of NVIDIA’s Grace and Hopper technology, power supercomputers built by major manufacturers like Dell and HPE. These chips offer a combined 200 exaflops (one quintillion floating point operations per second) of AI performance. Prominent computing centers, including Germany’s Jülich Supercomputing Centre and the University of Bristol, are integrating these superchips into their systems.

From decrypt

Gen AI for the Genome: LLM Predicts Characteristics of COVID Variants

Researchers from Argonne National Laboratory, NVIDIA, the University of Chicago, and other academic and commercial partners have harnessed the power of Generative AI to predict the characteristics of COVID-19 variants. Their model, known as GenSLMs, has demonstrated remarkable accuracy in generating gene sequences that closely resemble real-world variants of SARS-CoV-2, the virus responsible for COVID-19.

What sets GenSLMs apart is its ability to interpret long strings of nucleotides, akin to how a large language model (LLM) trained on English text would analyze a sentence. In this case, the nucleotide sequences represent the building blocks of DNA and RNA. Despite being trained solely on COVID-19 virus genomes from the first year of the pandemic, GenSLMs can predict gene mutations present in recent COVID strains, including the Eris and Pirola subvariants that have emerged this year. Remarkably, the model’s generative process lacks specific information or constraints about what a new COVID variant should look like, making its predictive capabilities even more impressive.

Beyond generating sequences, GenSLMs can classify and cluster different COVID genome sequences, distinguishing between variants. Users will soon be able to explore visualizations of GenSLMs’ analysis of evolutionary patterns within the COVID viral genome. By understanding how different parts of the genome co-evolve, researchers gain insights into potential vulnerabilities and forms of resistance that the virus may develop.

From nvidia

AI Faces Look More Real Than Actual Human Faces: A Surprising Study

AI has taken a leap into the uncanny valley, where its generated faces now appear more authentic than those of real people. Researchers from The Australian National University (ANU) have delved into this phenomenon, revealing intriguing insights.

The study, published in Psychological Science, focused on white faces produced by the popular StyleGAN2 algorithm. Participants were shown a mix of real human faces and AI-generated counterparts. Their task? To determine whether each face was genuine or crafted by AI. The results were astonishing: on average, people misjudged two out of three AI-generated faces as human. This phenomenon has been aptly dubbed the “hyperrealism” effect.

What’s even more intriguing is that our confidence in detecting AI impostors doesn’t align with our actual abilities. Participants who were least adept at spotting AI-generated faces paradoxically felt the most confident in their judgments. It’s as if the AI’s hyperrealism blindsided them, leaving them blissfully unaware of the deception.

From medicalxpress

OpenAI Tempts Google AI Talent with $10 Million Compensation Packages

In a high-stakes battle for top-tier talent, OpenAI has set its sights on luring seasoned researchers away from Google. According to a recent report by The Information, OpenAI is offering eye-popping pay packets—ranging from $5 million to $10 million—to entice Google’s best and brightest in the field of artificial intelligence.

But it’s not just about the money. OpenAI is also leveraging its reputation and resources to sweeten the deal. The company has already snagged dozens of former Google and Meta employees to bolster its ranks. These seasoned professionals are joining forces with OpenAI’s existing team to push the boundaries of AI research and development.

From businessinsider

Nov 12, 2023

Google in Advanced Talks to Invest in AI Startup Character.AI

Alphabet’s Google is currently engaged in discussions to invest hundreds of millions of dollars in Character.AI, an emerging artificial intelligence chatbot startup.

Founded by former Google employees Noam Shazeer and Daniel De Freitas, Character.AI has developed a generative AI tool that allows users to interact with virtual versions of famous individuals, both living and deceased.

These chatbots offer various roles and tones, catering to users aged 18 to 24, who contribute significantly to the platform’s website traffic. While Character.AI is free to use, it also offers a subscription model for users who want to skip the virtual line and access chatbots more quickly.

The potential investment from Google aims to support Character.AI’s model training efforts and meet the growing demand for its services. The talks are ongoing, and the terms of the deal may evolve.

From calcalistech

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.