AI Weekly Digest Issue #38

Get your AI update with AI Weekly Digest Issue #38 for Week 1 of 2024. A curated selection of the most impactful AI news and events worldwide.

Welcome back to AI Weekly Digest #38, your go-to resource for the most important artificial intelligence news and events from Week 1 of 2024.

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.

Subscribe to our newsletter and get the top 10 AI tools and apps delivered straight to your inbox. Subscribe now!

Jan 05, 2024

Table Of Contents

AI-powered search engine Perplexity AI, now valued at $520M, raises $73.6M
Perplexity AI

Perplexity AI, an AI-powered search engine, has raised $73.6M in a funding round led by Databricks Ventures and other investors, bringing its valuation to $520M .

The company, which was founded in 2022 by engineers with backgrounds in back-end systems, AI, and machine learning, offers a chatbot-like interface that allows users to ask questions in natural language, to which its AI responds by citing sites and sources from around the web.

Unlike a typical search engine, Perplexity offers a conversational answer engine that’s grounded in providing accurate and relevant information through citations ¹. The company is betting heavily on its AI tech to set it apart from the incumbents.

Perplexity reports that it had 10 million monthly visits and 2 million unique visitors in February alone. The fresh tranche comes as Perplexity reports that it had 10 million monthly visits and 2 million unique visitors in February alone.

Perplexity’s founders claim that their tech performs better, overall, when it comes to accuracy. With this new funding, Perplexity plans to expand its team and further develop its AI technology.

From techcrunch


AI Prescription Tool Backed by Mayo Clinic and Eli Lilly
Drug

UpDoc, a Palo Alto startup, is launching a voice-activated AI prescription management tool with backing from Mayo Clinic and Eli Lilly. The app, set to launch later this year at UCSF and Stanford, allows patients with chronic conditions like diabetes to interact with an AI assistant via smartphone to help monitor health and intervene when needed between doctor visits.

In a JAMA study of 32 diabetes patients, 81% maintained healthy glucose levels using UpDoc versus 25% without. The goal is to improve chronic care amid provider shortages. Doctors preset care plans, with humans intervening when patients enter gray areas. Led by Stanford physicians, UpDoc’s strategic funding round was led by Polaris Partners and includes Microsoft and Google as partners.

From axios


Google is working on a new AI-powered chatbot called Bard Advanced
Google Bard Gets Smarter by Connecting to Your Google Apps

Google is reportedly working on a new AI-powered chatbot called Bard Advanced, which is expected to rival OpenAI’s ChatGPT. According to a recent leak, Bard Advanced will be available through a Google One subscription and may introduce new features.

The chatbot is built on Google’s existing large language model, Lamda, which is known for its human-like responses. The company has not yet announced a release date for Bard Advanced, but the leak suggests that it could be launched soon.

From techradar


DeepMind’s Breakthrough in Advanced Robotics: AutoRT, SARA-RT, and RT-Trajectory
Google DeepMind

Google DeepMind’s latest blog post unveils significant advancements in robotics research with their new systems: AutoRT, SARA-RT, and RT-Trajectory. These innovations are set to enhance real-world robot data collection, decision-making speed, and environmental understanding.

AutoRT leverages large foundation models for practical training, while SARA-RT streamlines Robotics Transformer models for efficiency. RT-Trajectory focuses on helping robots generalize tasks using visual outlines.

From DeepMind


Meta and UC Berkeley Introduced A New Method For Creating Photorealistic Avatars From Audio Of Dyadic Conversations
From Audio to Photoreal Embodiment

Meta Reality Labs Research and the University of California, Berkeley have introduced an innovative method for creating photorealistic avatars from audio of dyadic conversations.

This technology synthesizes full-bodied human avatars, complete with facial expressions, body, and hand gestures, all driven by the dynamics of conversation audio.

Their method combines vector quantization and diffusion processes for more dynamic and expressive motions.

This breakthrough opens up new possibilities in digital communication, offering a more realistic and immersive virtual interaction experience.

From berkeley


OpenAI Offers Up To $5 Million For News Content Deals
OpenAI Screen

OpenAI is offering news publishers between $1-5 million per year to license copyrighted articles to train its AI models like ChatGPT, according to a report by The Information. This gives an indication of how much AI companies plan to pay for licensed news content.

It is similar to past non-AI licensing deals like Meta’s $3 million per year offer for Facebook News. But it is lower than bigger deals like Google’s $1 billion investment in news partnerships.

From theinformation


Jan 04, 2024

Google outlines new methods for training robots with video and large language models
AI Robot

Google’s DeepMind Robotics researchers are exploring new methods to give robots a better understanding of what humans want out of them. They are highlighting ongoing research designed to train robots with video and large language models.

The researchers are using large language models (LLMs) to help robots use tools more creatively. The rise of LLMs has tremendously enhanced the functionalities of chatbots, coding automation, and visual content creation.

Beyond these digital interfaces, embodied AI could represent the next frontier in intelligence—one that interacts tangibly with the real world. Robots, serving as the physical extensions of LLMs, present an ideal medium for this exploration.

From techcrunch


Camera Companies Add AI-Detecting Watermarks to Photos
Camera Companies Add AI-Detecting Watermarks to Photos
source: Canon

Nikon, Sony, and Canon recently announced a new initiative to embed digital watermarks in photos taken by their high-end mirrorless cameras. These watermarks will include metadata like time, location, and photographer details to cryptographically certify the authenticity of images. The goal is to help media professionals and photojournalists combat the spread of AI-generated fake images online.

From decrypt


Study Finds ChatGPT Has High Diagnostic Error Rate in Pediatric Case Diagnosis
Drug

According to a recent study published in JAMA Pediatrics, the popular AI chatbot ChatGPT had a diagnostic error rate of more than 80% in pediatric case diagnosis¹. The study analyzed the use of artificial intelligence in pediatric case diagnosis and found that ChatGPT had a high rate of diagnostic errors. The study’s findings suggest that the use of AI in pediatric case diagnosis should be approached with caution.

The study’s results are concerning, as ChatGPT is widely used in the healthcare industry to assist with patient diagnosis. The study’s authors suggest that further research is needed to determine the best way to use AI in pediatric case diagnosis.

From thehill


JPMorgan Announces DocLLM for Enterprise Document Analysis
JPMorgan Announces DocLLM

JPMorgan’s innovative approach to document analysis is making waves with their new generative language model, DocLLM. This groundbreaking tool is designed to navigate the complex landscapes of enterprise documents like forms, invoices, receipts, and contracts, which often blend textual and spatial information.

Unlike conventional models, DocLLM eschews heavy image encoders, focusing on bounding box data to comprehend spatial layouts. Its unique method decomposes attention mechanisms in transformers into disentangled matrices, adeptly handling both text and layout.

The model, pre-trained to infill text segments, excels in managing irregular layouts and diverse content. After fine-tuning with a vast instruction dataset, DocLLM has outshone state-of-the-art language models in 14 out of 16 datasets across four major document intelligence tasks, showing exceptional adaptability to new datasets.

From JPMorgan


Open AI will launch the GPT Store next week
OpenAI GPTs

OpenAI will launch the GPT Store next week. If you’re interested in sharing your GPT in the store, you’ll need to:

  • Review the updated usage policies and GPT brand guidelines to ensure that your GPT is compliant
  • Verify your Builder Profile (settings > builder profile > enable your name or a verified website) 
  • Publish your GPT as ‘Public’ (GPT’s with ‘Anyone with a link’ selected will not be shown in the store)

From OpenAI


AI Reveals Hidden Human Activity Across the Oceans
AI Reveals Hidden Human Activity Across the Oceans

Advances in AI and satellite technology have allowed researchers to create the most detailed global maps yet of human activity across the oceans. A new study led by nonprofit Global Fishing Watch analyzed five years of satellite imagery, exposing huge amounts of previously undetected fishing, transport, and offshore energy development.

The maps found that 75% of industrial fishing vessels are not publicly tracked, with hotspots of unmonitored activity off Africa and Asia. The data also revealed surges in offshore wind and oil platforms that were invisible to traditional tracking systems.

According to the researchers, the new AI techniques fill major blind spots in monitoring the oceans, providing vital information for global conservation efforts. With busy seas getting more crowded, the maps offer an unprecedented look at human impacts to guide protection of the marine environment.

From theverge


AI’s Existential Risk: A 5% Chance of Human Extinction
AI Robot

In a comprehensive survey of AI researchers, a notable consensus has emerged regarding the existential risk posed by the development of superhuman AI. The survey, the largest of its kind, included 2,700 researchers from top AI conferences. Findings show that 58% of the respondents see a 5% chance of AI leading to human extinction or similarly catastrophic outcomes. This perspective underscores the growing awareness within the AI community of the potential risks associated with advanced AI developments.

From newscientist


Microsoft adds AI button to keyboards to call up chatbot
Microsoft Unveils AI Assistant Copilot

Microsoft said it is adding a button to its Windows keyboard that would pull up the tech giant’s Copilot service, in the first change for the device in nearly three decades.

From Yahoo


Jan 03, 2024

Early Mickey Mouse is now in the public domain—and AI is already on the case
Mickey Mouse

On January 1, 2024, three early Mickey Mouse cartoons entered the public domain in the US. This has led to a surge in AI experiments that are taking advantage of this new development. Pierre-Carl Langlais, a digital humanities researcher, has uploaded an AI model to Hugging Face that has been trained on those public domain cartoons.

From arstechnica


Intel and DigitalBridge Unveil Articul8: A Leap into Generative AI for Business
Intel and DigitalBridge Launch Articul8

Intel Corp. and DigitalBridge Group Inc. have announced the launch of Articul8 AI Inc., an independent company focusing on generative AI (GenAI) for enterprises.

Articul8, leveraging technology from Intel, offers a full-stack, secure GenAI software platform designed for enterprise needs. The platform ensures data security and provides deployment flexibility, including cloud, on-premises, or hybrid options.

Arun Subramaniyan, former Intel executive, leads Articul8 as CEO. The initiative has garnered investment from Intel and several venture firms, aiming to integrate GenAI into various industries with high security and domain-specific requirements.

From intel


Mozilla Pivots to AI as Firefox Fades
Mozilla Firefox.jpg

Mozilla is shifting its focus to artificial intelligence as its Firefox browser continues to lose market share. In its 2023 corporate manifesto, Mozilla outlined plans to shape the internet’s future with open data and AI services. This represents a pivot away from Firefox, which is barely mentioned in the document despite being Mozilla’s main offering.

Meanwhile, Mozilla’s revenues dropped between 2021 and 2022 as Firefox’s market share declined from 3.79% to 3.04%. Yet CEO Mitchell Baker received a $1.3 million pay raise, now earning $6.9 million. Some suggest Mozilla intends to move beyond Firefox entirely.

Mozilla is building its financial assets as it eyes “difficult choices” in shutting down unprofitable projects. The foundation aims to develop trustworthy AI and machine learning to compete in this growing market. Mozilla seeks to balance people and profit while taking more risks. So while Firefox fades, Mozilla is staking its future on an open vision for artificial intelligence.

From techspot


MyShell releases OpenVoice voice cloning AI
Audio Voice

MyShell, a Canadian startup, recently launched OpenVoice, an innovative open-source AI for voice cloning.

This AI, developed with contributions from MIT and Tsinghua University, showcases remarkable efficiency and accuracy. It requires only a few seconds of audio to clone a voice, offering extensive control over aspects like tone, emotion, accent, rhythm, and more.

OpenVoice integrates two AI models: one for language style, including accents and emotions, and another for tone conversion. It has been trained on a vast array of audio samples in multiple languages.

From artificialintelligence-news


Midjourney Ventures into AI Video Creation
Midjourney launches its first mobile app, Niji Journey

The popular AI art generator Midjourney announced plans to expand into video creation. In an online session, CEO David Holz revealed that the company will begin training video models in January. The goal is to release text-to-video capabilities in a few months. This represents a natural progression for Midjourney as it leverages its mature image model to enter the burgeoning field of AI video generation

From decrypt


Jan 02, 2024

Nobel Prize Winner Warns Against Overemphasis on STEM Education in the Age of AI
child

Nobel Prize-winning economist, Paul Krugman, has warned against the overemphasis on science, technology, engineering, and mathematics (STEM) education in the age of artificial intelligence (AI).

Krugman argues that while STEM subjects are important, they are not the only skills that will be in demand in the future.

He suggests that the rise of AI will create a need for more empathetic and creative skills, which may not be as easily taught in a traditional STEM curriculum.

From Yahoo


Samsung’s Galaxy Unpacked 2024 Will Reveal New Mobile Experience Powered By AI
Samsung Galaxy AI

At Samsung’s Galaxy Unpacked 2024 event on January 17th, the company is poised to reveal a groundbreaking new lineup of Galaxy S series devices that promise to bring mobile AI to the next level. Powered by advanced on-device AI, the new Galaxy S series aims to deliver Samsung’s most intelligent smartphone experience yet, opening up possibilities for how people live, connect and create on the go.

From samsung


OpenAI Seeks Lead EU Privacy Oversight in Ireland
OpenAI Screen

OpenAI is moving European user data oversight to its Irish subsidiary ahead of a February 2024 update, likely seeking lead EU privacy regulation under Ireland’s Data Protection Commission. This would streamline GDPR oversight and reduce other EU states’ unilateral abilities to act on privacy concerns about products like ChatGPT.

From techcrunch


AI Discovers First New Antibiotic in Decades to Combat Drug-Resistant Infections
Health DNA

Scientists have used artificial intelligence to discover a new antibiotic capable of killing drug-resistant staph infections. The new compound is the first new antibiotic identified in over 60 years and represents a major breakthrough in the fight against antimicrobial resistance.

Researchers at MIT trained a deep learning model on large datasets to predict antibiotic activity against methicillin-resistant Staphylococcus aureus. By screening 12 million compounds, the AI model identified molecules from 5 classes predicted to combat the bacteria with low toxicity. Lab tests confirmed 2 promising candidates that reduced MRSA levels in mouse models.

From euronews


Ezra Makes AI-assisted Full-Body MRI Scans More Accessible for Earlier Cancer Detection
cancer medical

Ezra, a New York-based startup, has launched a full-body MRI screening service that can detect potential cancer and over 500 other conditions in up to 13 organs. The service is designed to help detect cancer early without harmful radiation. Early detection boosts cancer survival rates from 20% to 80%, the company says.

From decrypt


Meta’s 10 Most Interesting AI Research Advancements In 2023

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.

Leave a Reply

Your email address will not be published. Required fields are marked *

Get the latest & top AI tools sent directly to your email.

Subscribe now to explore the latest & top AI tools and resources, all in one convenient newsletter. No spam, we promise!