AI Weekly Digest Issue #34

Get your AI update with AI Weekly Digest Issue #34 for Week 49 of 2023. A curated selection of the most impactful AI news and events worldwide.

Welcome back to AI Weekly Digest #34, your go-to resource for the most important artificial intelligence news and events from Week 49 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.

Dec 08, 2023

Table Of Contents

E.U. reaches deal on landmark AI bill
European Union

The European Union has reached a landmark deal on the world’s most ambitious law to regulate artificial intelligence, paving the way for what could become a global standard to classify risk. The new law will require companies to provide detailed documentation of how their AI systems work, including the data they use to train them, and to ensure that their systems are transparent and auditable. The law will also require companies to provide clear explanations of how their AI systems make decisions, and to ensure that their systems are not biased against certain groups of people. The EU’s new law is seen as a major step forward in the regulation of AI, and is expected to set a global standard for the technology. The United States, by contrast, has yet to pass any significant legislation regulating AI, and is seen as lagging behind the EU in this area. The EU’s new law is expected to come into effect in 2025.

From washingtonpost


Nudify’ Apps That Use AI to Undress Women in Photos Are Soaring in Popularity, Prompting Worries About Non-Consensual Porn
Pika Video AI

The use of AI-powered apps that undress women in photos, known as “nudify” apps, is on the rise. These apps use artificial intelligence to remove clothing from images of people, making them appear naked. The development and rise of these apps have been met with excitement and curiosity. Some people are interested in the technology behind these apps, while others are curious to see what they can do.

However, the use of these apps has also raised concerns about non-consensual porn and the potential for abuse. Critics argue that these apps could be used to create fake nude images of people without their consent, which could then be shared online. This could lead to serious consequences for the victims, including harassment, bullying, and even job loss. As such, there is a growing need for regulation and oversight of these apps to ensure that they are used ethically and responsibly.

From fortune


NotebookLM: AI-Enhanced Note Management Now in the US
Google NotebookLM

Google’s AI-powered note-taking app, NotebookLM, is finally available in the US. This groundbreaking tool promises to revolutionize the way we capture and organize information, boosting productivity and unleashing creative potential.

NotebookLM leverages the power of artificial intelligence to provide a smarter note-taking experience. Unlike traditional note-taking apps, NotebookLM goes beyond simple text capture. It helps you understand and organize your notes, identify key insights, and even suggest next steps based on your content. This is all thanks to its sophisticated AI engine, trained on a massive dataset of text and code.

From Google


Google just launched a new AI and has already admitted at least one demo wasn’t real
Google Gemini Model

Google has recently launched a new AI called Gemini, which is designed to help users with their daily tasks. However, the company has already admitted that at least one of the demos shown during the launch event wasn’t real. The demo in question showed Gemini making a phone call to book a restaurant reservation, but it was later revealed that the call was pre-recorded and not made in real-time. This has led to accusations of misrepresentation, with some people questioning whether the other demos shown during the event were also staged.


Dec 07, 2023

Anthropic released a new dataset for measuring discrimination across 70 different potential applications of language models

AMD releases new chips to power faster AI training
AMD

AMD has announced the availability of new accelerators, the MI300A and MI300B, which are designed to power faster processing and model training for AI applications. The MI300A offers higher-performance computing, while the MI300B is optimized for generative AI workloads. The MI300A is equipped with 16GB of HBM2 memory and 7,680 stream processors, while the MI300B has 32GB of HBM2 memory and 15,360 stream processors.

Both accelerators are built on AMD’s CDNA 2 architecture and are compatible with the ROCm open-source software platform. The MI300A is priced at $7,000, while the MI300B is priced at $10,000. AMD’s new chips are expected to compete with Nvidia’s AI chips, which have been the industry standard for several years. The new accelerators are expected to be used in a variety of applications, including autonomous vehicles, medical imaging, and natural language processing.

From theverge


Meta’s AI image generator is now accessible via the web
Imagine with Meta AI Result

Imagine is Meta’s text-to-image generation service for the web. It allows users to turn text descriptions into high-quality images through Meta’s Emu image generation model.

Imagine with Meta AI serves as a bridge between imagination and visual art. By simply typing text prompts, users can bring their ideas to life in the form of detailed images. This free tool is ideal for graphic designers, content creators, and educators, providing a quick and efficient way to generate visuals for projects, presentations, or educational materials.

From Meta


Dec 06, 2023

Google Unveils Gemini: Their Largest and Most Capable AI Model
Google Gemini Model

Google has just unveiled its latest and most powerful AI model yet – Gemini. This groundbreaking development promises to reshape the AI landscape and unlock incredible possibilities across various industries.

From Google


Bard gets its biggest upgrade yet with Gemini
Google Bard & Gemini

Google has announced a significant upgrade to Bard, its conversational AI, by integrating the sophisticated Gemini AI model. Gemini, available in three sizes (Ultra, Pro, Nano), excels in multimodal reasoning and will enhance Bard’s capabilities in understanding, summarizing, reasoning, coding, and planning. The update begins with Gemini Pro, which has outperformed other models in benchmarks, and will soon include Gemini Ultra in Bard Advanced, offering advanced functionalities like processing text, images, audio, video, and code.

From Google


Google unveils AlphaCode 2, powered by Gemini
Google AlphaCode 2 Gemini

Google’s DeepMind lab has introduced AlphaCode 2, an advanced version of its code-generating AI, now powered by the Gemini Pro model. This enhanced AI demonstrates superior performance in programming competitions, excelling in languages like Python, Java, C++, and Go. AlphaCode 2 notably solves 43% of problems, nearly double its predecessor’s rate. It uses sophisticated techniques, including dynamic programming, to tackle complex mathematical and theoretical computer science challenges. While still facing challenges like operational costs, AlphaCode 2’s integration with Gemini suggests significant potential in AI-assisted programming.

From Techcrunch


Apple’s MLX: A New Contender in AI’s Open Source Landscape
Apple Inc

Apple’s foray into the open-source AI arena with MLX, a new deep-learning framework for Apple silicon, is a significant leap in machine learning technology. MLX, designed by Apple’s machine learning research team, stands out for its array framework specifically tailored for Apple silicon.

Key features of MLX include familiar APIs akin to NumPy and PyTorch, ensuring ease of use for developers. The framework boasts dynamic graph construction, allowing efficient model training without cumbersome compilation processes. Additionally, MLX supports lazy computation and multi-device operations, making it versatile and efficient.

From Apple


UK’s Bold Move: AI to Verify Age for Online Porn Access
UK

The UK has taken a significant step to safeguard minors from online pornography with the introduction of new age-check guidelines, prominently featuring AI technology. Under the Online Safety Act, websites and apps with pornographic content must prevent children from accessing their services. A disturbing study by the Office of the Children’s Commissioner for England revealed that children are exposed to online porn as early as age 9.

To address this, the government suggests using AI for facial age estimation, which would involve users uploading a selfie for age verification. Additional methods include photo ID verification and credit card checks. Open banking is another innovative approach, allowing users to authorize their banks to confirm their age to websites.

From reuters


ChatGPT Tops Wikipedia’s 2023 Viewership, Signaling a Surge in AI Interest
Most-viewed Wikipedia topics of 2023

In a remarkable testament to the growing influence of artificial intelligence, ChatGPT emerged as the most-viewed article on Wikipedia in 2023. This milestone underscores the burgeoning public interest and curiosity surrounding AI technologies.

From Axios


Dec 05, 2023

AMD Prepares to Challenge NVIDIA’s AI Chip Dominance with Wednesday Launch
AMD

Get ready for a shakeup in the artificial intelligence (AI) chip market, as AMD gears up for the Wednesday launch of their latest chip designed specifically for AI systems. This move marks a significant step in AMD’s strategy to compete with current industry leader, NVIDIA, and could potentially disrupt the current landscape.

While details on the chip’s specifications remain under wraps until Wednesday’s official announcement, analysts anticipate performance comparable to or even exceeding NVIDIA’s offerings. This could make AMD a viable alternative for companies developing AI applications, particularly those seeking cost-effective solutions.

From investopedia


Elon Musk’s AI firm xAI files to raise up to $1 billion in equity offering
xAI

Elon Musk’s latest venture in artificial intelligence, xAI, is making strategic financial moves by filing with the U.S. Securities and Exchange Commission to raise up to $1 billion through an equity offering. Already securing $134.7 million, xAI’s funding initiative underscores the burgeoning interest and investment in AI technologies, a trend gaining momentum since OpenAI’s launch of the groundbreaking ChatGPT and its significant backing by Microsoft.

From Yahoo


Tech Titans Unite: AI Alliance Launches to Champion Open, Safe, Responsible Artificial Intelligence
AI Alliance

IBM and Meta, along with over 50 other global collaborators, have launched the AI Alliance. This international community of technology leaders, researchers, and adopters aims to advance AI in an open, safe, and responsible manner. The Alliance focuses on harnessing AI advancements for improving work, life, learning, and interaction, emphasizing the importance of open and transparent innovation for safety, diversity, and economic opportunity.

The AI Alliance is action-oriented, bringing together a diverse group of organizations from various sectors to foster open innovation and science in AI. The goal is to create opportunities that reflect the complex needs of societies worldwide. By pooling resources and knowledge, the Alliance intends to address safety concerns and develop solutions for the global AI community.

From thealliance


Navigating the Tightrope: OpenAI and Amazon Tackle Chatbot Security Breaches
OpenAI Screen

In a recent development, OpenAI has addressed a significant security concern. Researchers discovered that by prompting ChatGPT to repeat a word indefinitely, it inadvertently disclosed internal data from OpenAI, including emails and phone numbers. This method was classified as a violation of service terms due to its resemblance to spamming. OpenAI has since amended its policies, making such attempts futile with warnings of content policy or terms of use violations.

Similarly, Amazon’s new AI agent, Q, faced scrutiny for over-sharing. A report highlighted that internal employee feedback, mistaken for a security breach, was actually standard practice for Amazon’s product development. Despite these challenges, both OpenAI and Amazon remain committed to enhancing their AI offerings, demonstrating the ongoing struggle to balance innovation with security in the rapidly evolving AI landscape.

From decrypt


AI Takes Over “Word of the Year” Lists, From “Hallucinate” to “Prompt”

Artificial intelligence has permeated every aspect of our lives, and now, it’s even influencing our language. In a surprising turn of events, “hallucinate,” “authentic,” and “prompt” have all emerged as contenders for “word of the year” in various prestigious dictionaries.

Cambridge Dictionary chose “hallucinate,” highlighting the tendency of AI models like ChatGPT to fabricate information rather than admitting their limitations. This raises concerns about misinformation and the blurred lines between reality and AI-generated content.

Merriam-Webster, on the other hand, focused on the impact of AI on authenticity with their selection of “authentic.” As deepfakes and AI-generated content become increasingly sophisticated, the question of what constitutes “real” becomes more pressing than ever.

Oxford University Press went a different route with “prompt,” reflecting the rise of generative AI and the human element involved in guiding them. This term emphasizes the collaborative nature of AI technology and the importance of human input.

While the “cooler terms” like “latent space” haven’t quite made the cut yet, it’s clear that AI is leaving an undeniable mark on our language. These “words of the year” are not just linguistic curiosities, but rather a reflection of the profound impact AI is having on our society. As AI continues to evolve, we can expect to see even more innovative and thought-provoking vocabulary emerge in the years to come.

From techcrunch


Dec 04, 2023

Microsoft’s Seeing AI App Now Available on Android, Making Visual Assistance More Accessible
Microsoft Seeing AI

Microsoft has expanded its Seeing AI mobile application designed to assist blind and low-vision users to Android devices. The app narrates the user’s surroundings to help them carry out everyday tasks like reading text, identifying objects, and describing photos or scenes. Key features include reading short and handwritten text aloud, recognizing currency, identifying people, summarizing articles, and answering questions about scanned documents.

From cnet


Anthropic Added Enhanced Controls For Claude Pro Users

Microsoft Unveils AI for Zero-shot Talking Avatar Generation

Microsoft has introduced GAIA, an AI system that can generate natural-looking talking avatar videos from just a single portrait photo and an audio clip. GAIA works by separating each video frame into motion and appearance parts. It then generates lifelike motion sequences based on the input speech, while preserving the appearance details from the reference portrait.

In tests, GAIA achieved superior realism, diversity, lip synchronization, and overall video quality compared to previous avatar creation methods. It is also highly scalable, with model performance improving alongside parameter count increases into the billions. Beyond talking avatars, GAIA enables pose-controllable avatar animation and text-driven avatar motion generation.

From Microsoft


Alibaba Unveils AI System for Animating Characters from Still Images

Alibaba has developed a new AI system called “Animate Anyone” that can generate consistent and controllable videos of animated characters from single image inputs. The system leverages diffusion models, which have shown promise for robust image and video generation. A key innovation is the ability to maintain detailed visual features and consistency across video frames, while allowing control over a character’s motion. This is enabled by components including a ReferenceNet module that merges in detail features from the reference image, an efficient pose guider that directs movements, and temporal modeling for smooth transitions between frames.

From Alibaba


OpenAI Files Trademarks for GPT-6 and GPT-7 in China
OpenAI Screen

As artificial intelligence firm OpenAI continues expanding the capabilities of its large language models, it has filed trademark applications in China for “GPT-6” and “GPT-7,” the next generations of its Generative Pre-trained Transformer models. Though none of OpenAI’s services are available in China, records show it submitted four Chinese trademark filings – two for GPT-6 and two for GPT-7 – which are still under review. This comes after OpenAI filed trademarks earlier this year for GPT-4 and GPT-5.

From scmp


Dec 03, 2023

OpenAI Bets on AI Chips From Altman-Backed Startup
OpenAI Screen

OpenAI, the artificial intelligence lab backed by Sam Altman, has committed to spending $51 million on specialized AI chips made by startup Rain Neuromorphics. As revealed in signed documents, OpenAI penned a letter of intent to purchase the brain-inspired chips over the next four years to power its AI systems.

The deal raises some eyebrows as Rain counts Altman as one of its earlier investors. However, an OpenAI spokesperson asserted Altman has recused himself from any decisions relating to purchases from Rain. The chips themselves aim to mimic the parallel processing of the human brain rather than the sequential processing of traditional computing. OpenAI expects the unconventional design to unlock new capabilities in training complex AI models. With Rain still working to commercialize its first chips, the OpenAI commitment offers a vital morale and financial boost to the young startup.


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 *