AI Weekly Digest Issue #35

Get your AI update with AI Weekly Digest Issue #35 for Week 50 of 2023. A curated selection of the most impactful AI news and events worldwide.

Welcome back to AI Weekly Digest #35, your go-to resource for the most important artificial intelligence news and events from Week 50 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 14, 2023

Table Of Contents

Stability AI Launches Tiered Membership for Enhanced AI Model Access
Stability AI Releases StableCode

Stability AI introduces a new membership system, offering tiered access to its cutting-edge AI models. The membership caters to a broad range of users, from individuals to large enterprises, and is structured into three tiers: Non-Commercial, Professional, and Enterprise. Each tier provides varying levels of access and benefits, tailored to suit different needs and applications. This initiative reflects Stability AI’s commitment to making advanced AI technology accessible while fostering innovation and creativity across various domains.

From Stability


DeepMind AI outdoes human mathematicians on unsolved problem
Google DeepMind

DeepMind’s AI system, FunSearch, has made significant strides in solving combinatorics problems inspired by the card game Set, surpassing human mathematicians’ efforts. This large language model-based system, described in Nature, automates the creation of short computer programs to generate novel solutions. It outperformed previous human findings in the cap set problem, a complex variant of Set. This breakthrough not only showcases AI’s potential in advancing mathematical understanding but also emphasizes the evolving synergy between human and machine intelligence.

From Nature


Revolutionizing AI Alignment: OpenAI’s New Approach in Weak-to-Strong Generalization
OpenAI Screen

OpenAI introduces a groundbreaking approach in AI alignment with their research on weak-to-strong generalization. The concept focuses on using less capable AI models to supervise more advanced ones. This methodology aims to tackle the challenge of aligning superintelligent AI systems, ensuring they remain beneficial and safe for humanity. OpenAI’s initial experiments demonstrate the potential of this approach, where a GPT-2-level model can effectively supervise GPT-4, achieving impressive results. This new direction in research could significantly impact the future development and control of superhuman AI systems.

From OpenAI


Intel unveils new AI chip to compete with Nvidia and AMD
Chip

Intel has announced the launch of its new AI chip, Gaudi3, which is designed to compete with Nvidia and AMD. The chip is specifically designed for generative AI software and is expected to provide a significant boost in performance. Intel has also announced the release of Core Ultra and 5th gen Xeon server chips. The Gaudi3 chip is expected to be a game-changer in the AI industry, as it will allow developers to create more complex and sophisticated AI models.

The chip is expected to be available in the market soon, and it is expected to be a significant competitor to Nvidia and AMD’s AI chips. Intel’s new AI chip is expected to be a significant step forward in the development of AI technology, and it is expected to have a significant impact on the AI industry. The Gaudi3 chip is expected to be a significant breakthrough in the development of AI technology, and it is expected to be a significant competitor to Nvidia and AMD’s AI chips. Stay tuned for more updates on this exciting development in the AI industry!

From cnbc


Instagram introduces gen-AI powered background editing tool
Instagram

Instagram has introduced a new background editing tool powered by generative AI¹. The tool is called “Backdrop” and allows users to replace the background in an image with a new one of their choice. Users can type prompts like “put me in front of a sublime aurora borealis” and the tool will create a new version of the image with the relevant background applied¹. This feature is expected to be available to Instagram users soon, although no specific date has been announced yet¹.

From techcrunch


Dec 13, 2023

Google Unveils Imagen 2: A New Era in Image Generation with Vertex AI
Google Unveils Imagen 2

Google Cloud has announced the general availability of Imagen 2 on Vertex AI, marking a significant advancement in image-generation technology. Developed with Google DeepMind, Imagen 2 offers high-quality, photorealistic images from textual prompts. It supports text rendering in multiple languages, logo generation, and visual question-answering capabilitie

From Google


Google Cloud Announced MedLM: A generative AI fine-tuned for the healthcare industry
Google Cloud MedLM

Google Cloud introduces MedLM, a specialized AI for the healthcare industry, now available in the U.S. on Vertex AI. MedLM, built on Med-PaLM 2, comes in two models tailored for healthcare tasks, aiming to improve various applications from medical note-taking to pre-clinical research. The larger model handles complex tasks, while the medium model is adaptable for scaling across tasks. Notable deployments include HCA Healthcare for medical documentation and BenchSci for enhancing pre-clinical R&D. MedLM’s integration with healthcare solutions by Accenture and Deloitte further highlights its potential in transforming healthcare processes and patient outcomes.

From Google


Google’s Gemini Pro is now accessible to developers and enterprises
Google Gemini Model

Google has launched Gemini Pro, a highly capable AI model now accessible to developers and enterprises. This new technology, available through the Gemini API, offers advanced features like a 32K text context window, multi-language support, and a special Vision multimodal endpoint. Gemini Pro promises enhanced app development capabilities with free access and SDKs for various platforms. It also integrates seamlessly with Google AI Studio and Vertex AI on Google Cloud, ensuring a smooth transition from development to enterprise-level deployment with robust security and privacy features.

From Google


Stability AI Introduces Stable Zero123: Quality 3D Object Generation from Single Images
Stability AI Stable Zero123

Stability AI introduces Stable Zero123, a breakthrough in 3D object generation from single images. Leveraging an enhanced training dataset, Stable Zero123 surpasses its predecessor, Zero123-XL, in creating realistic 3D views. This advancement is credited to improved dataset quality, elevation conditioning, and training efficiency. Intended for non-commercial and research use, this model is available on Hugging Face. It’s a significant step forward in 3D imaging technology, enabling more detailed and accurate reconstructions from 2D images.

From Stability


OpenAI and Axel Springer Forge Path in AI-Enhanced Journalism
OpenAI Screen

OpenAI and Axel Springer have announced a pioneering partnership to integrate AI in journalism. This collaboration will enrich ChatGPT with diverse, authoritative content from Axel Springer’s media brands, including POLITICO and BUSINESS INSIDER. It aims to elevate journalistic quality and societal relevance while exploring new financial opportunities for sustainable journalism. This partnership also involves leveraging Axel Springer’s content to advance the training of OpenAI’s language models, marking a significant step in the evolution of AI-powered news dissemination.

From OpenAI


OpenAI’s nonprofit arm showed revenue of $45,000 last year

OpenAI’s nonprofit arm, OpenAI LP, reported a revenue of $45,000 in 2022, according to the company’s tax filing. This is a significant drop from the $1.5 million revenue reported in 2021. Despite the drop in revenue, OpenAI LP’s parent company, OpenAI Inc., is worth billions of dollars.

From cnbc


Ray-Ban has partnered with Meta to create smart glasses
Sunglasses

Ray-Ban has partnered with Meta to create smart glasses that can identify objects and translate languages. The glasses are equipped with Meta’s multimodal AI, which allows them to recognize landmarks and translate signs in real-time. The glasses are also capable of livestreaming, making them a versatile tool for a variety of applications. While the glasses do not have a heads-up display, they can be controlled through voice commands, making them a hands-free option for users on the go. The glasses are expected to be available for purchase in the near future, and are sure to be a hit with anyone looking for a convenient and stylish way to stay connected and informed while on the move.

From theverge


Claude released a new prompt engineering tool that makes Claude accessible via spreadsheets

Dec 12, 2023

Pika Art is beginning to let people in off the waitlist

Microsoft Announces Phi-2: The surprising power of small language models
Microsoft

Microsoft Research has unveiled Phi-2, a 2.7 billion-parameter language model, notable for its compact size and powerful capabilities in reasoning and language understanding. Outperforming larger models, Phi-2 is a testament to strategic training data selection and innovative scaling techniques. It provides an ideal platform for researchers, allowing for exploration in various AI domains. Phi-2 is accessible in the Azure AI Studio model catalog, encouraging further research and development in the field.

From Microsoft


Snapchat now lets subscribers share AI-generated snaps
Snapchat

Snapchat has launched a new feature called “My AI Snaps” for its subscribers. The feature allows users to create and share AI-generated images. Subscribers can access the feature by tapping the “AI” button on the right side of the camera interface. Users can then type a text prompt or choose from one of the premade options, such as “a futuristic disco” or “a rocket preparing for liftoff.” The AI then generates an image based on the prompt.

The feature is only available to subscribers of Snapchat’s $3.99 plan. The company has also released other AI-powered features for its subscribers, including the ability to add the My AI chatbot into group chats with friends, where it can be summoned with an @ mention. The chatbot can also be used for place recommendations and lens suggestions. The My AI chatbot was recently updated to be able to reply to users’ Snaps with a text-based reply. With the paid subscription to Snapchat+, My AI can now return a photo as well.

From theverge


Sports Illustrated CEO Fired Following AI Controversy
Stability AI Releases StableCode

The Arena Group, owner of Sports Illustrated, announced the termination of its CEO Ross Levinsohn on Monday, following revelations that the magazine had been publishing articles generated by artificial intelligence. The decision came after a meeting of the board to discuss “improving the operational efficiency and revenue” of the company.

Levinsohn’s firing marks a significant development in the ongoing debate about the role of AI in journalism. While AI has the potential to revolutionize many industries, its use in creating news articles has raised ethical concerns about transparency and authenticity.

Levinsohn had previously touted AI’s ability to create enterprise value for brands, but the recent scandal has damaged Sports Illustrated’s reputation and cast doubt on the future of AI-generated journalism. It remains to be seen how the media industry will respond to this controversy and whether AI can regain its foothold in the newsroom.

From theguardian


Microsoft and Labor Unions Join Forces to Ensure Ethical Use of AI in the Workplace

Microsoft has formed an “historic” alliance with labor unions to address the use of artificial intelligence (AI) in the workplace. The agreement is aimed at ensuring that AI is used in a way that is ethical, transparent, and accountable, and that it benefits workers as well as employers. The partnership is the first of its kind in the United States and is expected to serve as a model for other companies and unions¹. The agreement includes provisions for training and reskilling workers, as well as for ensuring that workers have a say in how AI is used in their workplaces. The partnership is seen as a significant step forward in the effort to ensure that AI is used in a way that benefits everyone, not just a select few.

From Yahoo


EU begin to hash out EU AI Act details starting Tuesday
European Union

The EU is embarking on a pivotal journey to regulate artificial intelligence. After securing a political agreement, experts from EU countries and lawmakers are set to detail the AI Act’s provisions. Key discussions will focus on AI’s use in biometric surveillance, foundation model regulations, and major systems like ChatGPT. This marks a major step towards the EU becoming the first to legislate in the rapidly evolving AI sector. The finalized text, expected to take effect in 2026, will undergo formal editing before publication. Nations like France and Germany are keenly analyzing the compromises, highlighting the Act’s significance in shaping future AI governance.

From Reuters


Sam Altman has acknowledged that ChatGPT could potentially replace millions of jobs in the future
ChatGPT Unsplash

Sam Altman, CEO of OpenAI, has acknowledged that the company’s AI system, ChatGPT, could potentially replace millions of jobs in the future. Altman also noted that the chatbot is likely to change many other jobs in the process. While the transition to an AI-enabled future is mostly good and can happen fast, Altman cautioned that society needs time to adapt to something so big as AI. He also emphasized that regulation will be critical and will take time to figure out. Altman’s comments echo those of Mira Murati, the CTO of OpenAI, who said that ChatGPT should be regulated, as it could be misused.

From livemint


Mind-reading AI can translate brainwaves into written text
artificial-intelligence-brain

The New Scientist article discusses a breakthrough in AI technology where scientists can convert brainwaves into written text. This innovation uses a helmet with sensors and AI to interpret EEG recordings of brain activity as people read text. The current model, named DeWave, has shown an accuracy of around 40%, with newer data suggesting improvements above 60%. This method is more practical compared to MRI-based approaches, as it doesn’t require subjects to remain immobile. The technology, which pairs DeWave with a large language model, holds potential for aiding those who have lost speech ability and could also find applications in robotics.

From newscientist


Dec 11, 2023

Meta Announces Audiobox Demos
Meta Announces Audiobox Demos

Audiobox is Meta’s new foundation research model for audio generation. It can generate voices and sound effects using a combination of voice inputs and natural language text prompts — making it easy to create custom audio for a wide range of use cases. The Audiobox family of models also includes specialist models Audiobox Speech and Audiobox Sound, and all Audiobox models are built upon the shared self-supervised model Audiobox SSL.

From Meta


Investigating ChatGPT’s Alleged ‘Laziness’: OpenAI’s Response to User Concerns
ChatGPT New

Recent user reports suggest that ChatGPT, built on OpenAI’s GPT-4 model, has become less cooperative, leading OpenAI to investigate these claims. Users, primarily on Reddit and OpenAI’s forums, have noticed that ChatGPT sometimes provides minimal information or directs users to complete tasks themselves, behavior that some perceive as ‘lazy’ or ‘sassy’. Speculations arise that this might be a strategic move by OpenAI to optimize ChatGPT’s efficiency, considering the high operational costs of AI systems. OpenAI acknowledges these concerns but clarifies that no recent changes have been made to ChatGPT that would explain this behavior. The organization is exploring the issue, emphasizing the unpredictability of AI model behavior.

From independent


Advancing Transparency in AI Research through LLM360
LLM360

LLM360 is a new open-source initiative that aims to increase transparency and trust in large language models (LLMs) through comprehensive model releases. When publishing an LLM under LLM360, the goal is to provide full access to intermediate checkpoints, all training data mapped to checkpoints, logs and metrics, and source code. This additional information allows for deeper analysis into an LLM’s construction and behavior.

From LLM360


Elon Musk’s Grok Ai Caught Plagiarizing Openai’s Chatgpt
X AI Grok

Elon Musk’s AI startup, Grok, has been caught plagiarizing OpenAI’s ChatGPT. The chatbot, which was launched last week, was found to have copied text from the OpenAI model. This comes after Musk publicly criticized ChatGPT for being “woke” and training AI to lie.

From futurism


Mistral AI’s Strategic Leap: $415M Funding Fuels European AI Ambition
Mistral AI

Mistral AI, a Paris-based startup, has completed a significant Series A funding round, securing €385 million ($415 million), valuing it at around $2 billion. This round follows a $112 million seed round just six months prior. With a focus on foundational models in generative AI, Mistral AI aims to be a European rival to OpenAI, backed by notable investors including Andreessen Horowitz and Lightspeed Venture Partners.

The company’s approach is unique, offering its large language model, Mistral 7B, for free download under an open-source license, rather than through APIs. This model, developed with a “small” 7 billion parameter dataset, indicates a strategic shift towards open, responsible, and decentralized AI technology. Mistral AI also actively contributes to shaping the EU’s AI Act, advocating for regulations on AI use cases rather than foundational models.

From techcrunch


Microsoft’s Partnership With OpenAI Faces UK Antitrust Scrutiny
Microsoft

Microsoft’s partnership with OpenAI is under scrutiny by the UK antitrust watchdog, the Competition and Markets Authority (CMA), which is considering whether to launch a full-blown investigation. The partnership has recently gone through a governance meltdown and the firing of Sam Altman as chief of OpenAI last month exposed how inextricably linked the two companies have become.

Microsoft has invested as much as $13 billion in OpenAI, and by integrating OpenAI’s products into virtually every corner of its core businesses, the software giant has established itself as the undisputed leader of AI among big tech firms. The CMA is taking comments until Jan. 3 before making a decision on a formal investigation.

Microsoft President Brad Smith said in a statement that “the only thing that has changed is that Microsoft will now have a non-voting observer on OpenAI’s board.” He described its relationship with OpenAI as “very different” from Google’s outright acquisition of DeepMind in the UK. OpenAI didn’t immediately respond to a request for comment.

From Bloomberg


Spotify to cut 6% of its workforce to focus on AI
Spotify

Spotify has announced that it will be cutting its workforce by 6% or about 600 people, as it seeks to move faster into AI. The company is betting on AI to boost margins for its podcasting and audiobook divisions. The move has been welcomed by Wall Street, which seems confident that it can work.

Spotify’s CEO, Daniel Ek, said that the company’s growth of operating expenses outpaced its revenue growth by two times in 2022. He added that it would have been unsustainable long-term in any climate, but with a challenging macro environment, it would be even more difficult to close the gap. The company expects to take severance costs of between $38 million and $49 million in connection with the layoffs.

Dawn Ostroff, one of Spotify’s most high-profile executives, will depart the company “as part of a broader reorganization,” the company said. Ostroff will become a senior adviser to Spotify to help with the transition. Alex Norström, currently chief freemium business officer, and Gustav Söderström, currently chief research and development officer, are being elevated to co-presidents of the company. Norström will oversee content issues.

From CNN


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 *