Last Week in AI #239: Mistral AI's free LLM 🤖, Hollywood's AI deal 🎬, OpenAI faces another lawsuit from authors 📚, and a bipartisan 'light-touch' AI bill in the works 🏛️, and more!

Mistral AI releases highly performant LLM under Apache 2.0, film studios agree AI cannot replace writers, top authors sue OpenAI, light-touch AI regulation bill gaining traction

Top News

Mistral AI makes its first large language model free for everyone

French AI startup, Mistral, has released its first large language model, Mistral 7B, for free use without restrictions under the Apache 2.0 license. The model, which the company claims outperforms others of its size, is available for download via various means, including a 13.4-gigabyte torrent. Mistral 7B is a refined version of other "small" large language models, offering similar capabilities at a significantly lower compute cost. The company has also started a GitHub repository and Discord channel for collaboration and troubleshooting. While the model is free to use, it is not open source as it was developed privately with private funds, and the datasets and weights are also private. Mistral's business model appears to be offering the free model for use, but charging for their commercial offering, which includes both weights and code sources.

Hollywood’s Deal With Screenwriters Just Rewrote the Rules Around A.I.

The Writers Guild of America (W.G.A.) has negotiated a deal with Hollywood studios regarding the use of AI in scriptwriting. The agreement ensures that AI cannot replace credited and paid guild members, although it can be used to generate first drafts. The credited writers will receive the same minimum pay as if they had written the piece from scratch. The deal also addresses copyright concerns, as material generated by AI cannot be copyrighted. The W.G.A.'s approach mirrors successful labor negotiation tactics from the past, such as the International Longshore and Warehouse Union's acceptance of port containerization in 1960, which resulted in wage and job guarantees for most workers and benefits for those displaced.

Franzen, Grisham and Other Prominent Authors Sue OpenAI

Prominent authors including Jonathan Franzen, John Grisham, David Baldacci, Jodi Picoult, George R.R. Martin, George Saunders, and Michael Connelly have filed a lawsuit against OpenAI, represented by the Authors Guild. The lawsuit is in response to ChatGPT's ability to generate detailed information about their books, beyond what is publicly available, as demonstrated by novelist Douglas Preston's interaction with ChatGPT. Since its introduction in November, ChatGPT has been a point of contention in the publishing industry, with a surge in AI-generated books appearing on Amazon. In response, Amazon has implemented new guidelines for self-published authors, requiring them to disclose AI usage in text creation and limiting the number of titles they can upload daily.

Top GOP senator teams up with key Dem on ‘light-touch’ AI bill

Top GOP senator, Thune, and key Democrat, Klobuchar, are collaborating on a 'light-touch' AI bill aimed at mitigating the risks of AI applications while avoiding heavy regulation. The bill, which has not yet been officially introduced, has already gained support from influential figures such as Ryan Hagemann, AI policy executive at IBM. The proposed legislation would task the Commerce Department with enforcement, but companies developing or deploying AI systems would be responsible for assessing their impact and certifying their safety. The bill is expected to gain more Senate cosponsors ahead of its official introduction, which is anticipated to occur within the next week or two.

Other News

Applications

How California is using AI to snuff out wildfires before they explode - The California Department of Forestry and Fire Protection is using artificial intelligence to detect wildfires early and prevent them from spreading, with the technology successfully identifying 40% of fires since July 10 before a 911 call was received.

Google and the Department of Defense are building an AI-powered microscope to help doctors spot cancer - Google and the Department of Defense have developed an AI-powered microscope called the Augmented Reality Microscope (ARM) that can help pathologists diagnose cancer more accurately and efficiently, potentially serving as a second opinion for pathologists in remote areas or with limited access to expertise.

Oracle brings voice-activated AI to healthcare with Clinical Digital Assistant - Oracle has announced the integration of an AI-powered Clinical Digital Assistant into its healthcare solutions, aiming to automate administrative tasks and improve the quality of patient care.

Fighting Canada’s record wildfires with a combination of AI and intuition - AI is being used to accurately predict the likelihood of new wildfires in Canada, increasing duty officers' trust in the tool and potentially saving millions of dollars in operating costs.

AI is policing the package theft beat for UPS as 'porch piracy' surge continues across U.S. - UPS has introduced an AI-powered approach called DeliveryDefense to reduce the risk of package theft, using historic data and machine learning algorithms to assign each location a "delivery confidence score" and allowing package recipients to recommend in-store collection or a UPS pick-up point for low-scoring addresses.

Why Silicon Valley’s biggest AI developers are hiring poets - Silicon Valley's biggest AI developers are hiring poets and writers with humanities degrees to write short stories and provide feedback on the literary quality of AI-generated text, as they seek to improve the fluency and credibility of their generative writing products.

Spotify tests voice translation feature for podcasts - Spotify is testing an AI-powered feature that will translate podcasts to other languages, using OpenAI's voice generation technology, in an effort to attract new users and boost revenue.

Tubi is using ChatGPT to give you better movie recommendations - Tubi is using ChatGPT's Rabbit AI feature to provide users with conversational and AI-powered search capabilities for better movie recommendations, allowing users to find content based on various prompts and preferences.

Toyota is making AI-trained breakfast bots in a ‘kindergarten for robots’ - Toyota Research Institute (TRI) used generative AI in a “kindergarten for robots” to teach robots how to make breakfast by giving them a sense of touch and showing them how, allowing them to learn new behaviors overnight.

ChatGPT can now search the web in real time - ChatGPT can now search the web in real time, providing answers from current and authoritative sources, and allowing users to verify the information and give credit to the sources.

Google is opening up its generative AI search experience to teenagers - Google is opening up its generative AI search experience to teenagers, allowing them to ask questions in a conversational manner and adding safeguards to prevent inappropriate or harmful content from surfacing, while also introducing a new feature to provide users with more context about the content they see.

Business

China's AI 'war of a hundred models' heads for a shakeout - China's AI industry, driven by the success of OpenAI's ChatGPT, is experiencing a surge in product announcements, but investors warn that a shakeout is imminent due to rising costs and lack of viable business models.

A Silicon Valley Supergroup Is Coming Together to Create an A.I. Device - Sam Altman and Jony Ive are teaming up to develop a new AI device that would succeed the smartphone and deliver the benefits of AI in a new form factor, seeking funding from SoftBank to support the project.

AI Startup Writer Pens $100M Round - Generative AI platform Writer has raised $100 million in a funding round led by Iconiq Growth, valuing the startup at $500 million, as it continues to help businesses generate writing and content using large language models.

AI startup AlphaSense valued at $2.5 billion after latest funding round - Market intelligence platform AlphaSense has raised $150 million in a funding round, bringing its valuation to $2.5 billion, highlighting the high demand for AI startups.

Intel says newest laptop chips, software will handle generative AI - Intel's new laptop chip, set to be released in December, will allow generative AI chatbots to run on laptops without relying on cloud data centers, enabling businesses and consumers to test AI technologies without sending sensitive data off their own computer.

Amazon gives Alexa an AI facelift as it launches new smart speakers - Amazon introduces a "smarter and more conversational" version of its Alexa voice assistant, equipped with generative AI features, in an effort to bolster its position in the AI race.

Roblox acquires AI voice moderation start-up Speechly - Roblox has acquired Finnish AI start-up Speechly to enhance its voice chat service, with Speechly's technology being used to improve moderation on the platform.

Getty made an AI generator that only trained on its licensed images - Getty Images has partnered with Nvidia to launch Generative AI, a tool that allows users to create images using Getty's library of licensed photos, with the AI model trained exclusively on Getty's vast image collection to provide copyright indemnification.

Morgan Stanley debuts a new tool for employees: an AI assistant to answer common investing and personal finance queries - Morgan Stanley has launched an AI assistant, built from OpenAI's GPT-4 language model, to help financial advisors answer common investing and personal finance queries.

Cloudflare launches new AI tools to help customers deploy and run models - Cloudflare has launched a suite of AI tools, including Workers AI, Vectorize, and AI Gateway, to help customers build, deploy, and run AI models at the network edge, with a focus on cost savings and low-latency user experiences.

AWS announces the general availability of Amazon Bedrock and powerful new offerings to accelerate generative AI innovation - AWS has announced the general availability of Amazon Bedrock and other generative AI innovations, allowing organizations to build new generative AI applications and transform businesses.

Google discussed dropping Broadcom as AI chips supplier - The Information - Google has discussed dropping Broadcom as a supplier of AI chips, potentially saving billions of dollars in costs annually, according to a report by The Information.

Deci AI Unveils DeciDiffusion 1.0: A 820 Million Parameter Text-to-Image Latent Diffusion Model and 3x the Speed of Stable Diffusion - Deci AI introduces DeciDiffusion 1.0, a groundbreaking text-to-image generation model that produces high-quality images more efficiently, bridging the gap between natural language understanding and visual content creation.

Meet the A.I. Jane Austen: Meta Weaves A.I. Throughout Its Apps - Meta has introduced a suite of products powered by artificial intelligence, including an AI character based on Jane Austen, that will be integrated into its messaging apps and other platforms.

Bumble CEO Whitney Wolfe Herd shares how AI will ‘supercharge’ love and relationships - Bumble CEO Whitney Wolfe Herd believes that AI technology will enhance the online dating experience by improving matching algorithms, reducing friction, and offering a more curated and premium service.

Former Tinder CEO’s startup to fight loneliness with AI chatbot gets backing by Sequoia - Former Tinder CEO Renate Nyborg's startup Meeno, which aims to fight loneliness through an AI-powered chatbot, has raised $3.9 million in funding led by Sequoia and plans to release its app on the App Store in December in multiple countries.

Adobe launches Photoshop’s web version with Firefly-powered AI tools - Adobe has officially launched Photoshop for the web, offering Firefly-powered AI tools such as generative fill and generative expand, as well as collaboration features, to paid users.

Humane’s ‘AI Pin’ debuts on the Paris runway - Humane's 'AI Pin' wearable device, which promises to leverage AI in unique ways, made its debut on the Paris runway during Coperni's 2024 Spring Summer show at Paris Fashion Week, with supermodel Naomi Campbell being the first person outside of the company to wear the device in public.

Gates Foundation backs GE HealthCare AI-based ultrasound - GE HealthCare has received a $44 million grant from the Bill & Melinda Gates Foundation to develop user-friendly, AI-assisted ultrasound imaging tools that can provide clinical decision information and support more effective ultrasound scans, particularly in low- and middle-income countries.

Apple to buck layoff trend by hiring UK AI staff - Apple plans to hire more staff in the UK to work in artificial intelligence, in contrast to the trend of layoffs in the tech sector.

The Llama Ecosystem: Past, Present, and Future - The Llama Ecosystem, built on the belief in the power of open source, aims to make state-of-the-art AI technology more accessible and safer through collaboration with the research community, enterprise and commercialization, and the development of a developer ecosystem.

AI startup Lamini bets future on AMD's Instinct GPUs - AI startup Lamini is betting its future on AMD's Instinct GPUs, running its large language model (LLM) refining platform exclusively on the silicon, attracting interest from major companies like Amazon, Walmart, and Adobe.

Indeed’s “AI at Work Report” Finds GenAI will Impact Almost Every Job in America - Indeed's AI at Work Report reveals that GenAI (generative artificial intelligence) will impact almost every job in America, with software development jobs being the most exposed and driving jobs being the least exposed, but overall, GenAI is more likely to augment or streamline parts of a job rather than fully replace it.

Microsoft Will Roll Out Copilot AI Companion This Fall - Microsoft is rolling out its AI assistant, Copilot, across Windows 11, Bing, Edge, and Microsoft 365 this fall, aiming to provide context-sensitive assistance and a seamless experience across multiple products.

Concerns

Three in Four Americans Believe AI Will Reduce Jobs - Three in four Americans believe that artificial intelligence (AI) will reduce jobs, although they recognize the benefits of AI in certain areas and have low trust in businesses to use AI responsibly.

Third-party AI tools pose increasing risks for organizations - The increasing use of third-party AI tools by organizations poses risks such as reputational damage, financial losses, and regulatory penalties, highlighting the need for companies to develop responsible AI frameworks to mitigate these risks.

AI facial recognition tech leads to wave of lawsuits from Black plaintiffs after mistaken identities end in arrests - Black plaintiffs are filing lawsuits against law enforcement agencies after being mistakenly identified and arrested due to the misuse of facial recognition technology, with critics arguing that the technology results in a higher rate of misidentification of people of color than of white people.

AI voice cloning scams on the rise, expert warns - Scammers are increasingly using AI voice cloning tools to place panicked phone calls to individuals' family or friends, in an attempt to convince them to give money or sensitive information, according to an expert.

OpenAI and ChatGPT Lawsuit List - OpenAI is facing multiple lawsuits, including allegations of copyright infringement, privacy violations, and libel, which could have significant implications for the development and use of AI technologies.

First Google Search Result for Tiananmen Square “Tank Man” Is AI Generated Selfie - Google has removed an AI-generated selfie of the iconic "Tank Man" from its search results after it appeared as the top result, highlighting the challenge of identifying and differentiating AI-generated content from human-made content.

AI Book Bans: Testing LLMs Against the Freedom to Read - Large language models were tested to see how they would justify book bans, with some models complying with the request to remove books while others defended the freedom to read, highlighting the need for a "Librarianship of AI" to guide the behavior of AI models and ensure access to diverse perspectives.

Six Months Ago Elon Musk Called for a Pause on AI. Instead Development Sped Up - Six months ago, an open letter calling for a pause on AI development did not result in a slowdown, but it did help bring the idea of AI's potential dangers into the mainstream.

How the “Surveillance AI Pipeline” Literally Objectifies Human Beings - Computer vision research often leads to technology that surveils human beings, with computer vision papers frequently referring to humans as "objects," according to a new study that analyzed thousands of papers and patents spanning three decades.

Weizenbaum’s nightmares: how the inventor of the first chatbot turned against AI - Joseph Weizenbaum, the inventor of the first chatbot, became disillusioned with AI and warned of its dangers, believing that it strengthened repressive power structures and flattened human reason into senseless routines of code, leading to a world that was more unequal and less rational.

Tom Hanks warns dental plan ad image is AI fake - Tom Hanks warns about the prevalence of AI-generated fake images and discusses the challenges it poses to the entertainment industry.

Fake naked pictures of young girls created with AI spark fury in a small Spanish town - AI-generated fake naked pictures of young girls have caused outrage in a small Spanish town, with over 20 victims identified and at least 10 suspects under investigation.

Project Gutenberg has implemented one of the worst AI fears of striking actors - Project Gutenberg has used AI to transform thousands of ebooks into audiobooks, bypassing the need for human readers, which raises concerns for actors currently on strike who fear that AI may endanger their careers.

AI Is Defeated in Hollywood—But What About Music? - AI music is becoming more prevalent on streaming platforms, but it is often dull and unexciting, leading to concerns about the future of music culture and the need for transparency and quality in AI-generated music.

Can you melt eggs? Quora’s AI says “yes,” and Google is sharing the result - Google's Featured Snippet, which provides quick answers to search queries, is pulling incorrect information from Quora's AI feature, ChatGPT, resulting in false information being propagated and potentially eroding trust in Google's search results.

Research

AI system found to outperform humans in creating urban planning designs - An AI-based urban planning system has been found to outperform human experts in creating urban planning designs, according to a study by a team of urban planners and information scientists at Tsinghua University in China.

Chan Zuckerberg Initiative is building an expensive new AI GPU cluster for medical research - The Chan Zuckerberg Initiative is funding the construction of a massive computing system for medical research, consisting of over 1,000 top-of-the-line GPUs, with the aim of using generative AI to study healthy and diseased cells and make groundbreaking discoveries in the field of medicine.

DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation - DreamGaussian is a novel 3D content generation framework that achieves efficiency and quality for image-to-3D and text-to-3D tasks, producing high-quality textured meshes in just a few minutes.

Text-to-3D using Gaussian Splatting - A novel approach called Gaussian Splatting based text-to-3D generation (GSGEN) is presented in this paper, which leverages 3D Gaussian Splatting to generate high-quality 3D objects with accurate geometry and improved fidelity.

Kosmos-2.5: A Multimodal Literate Model - A multimodal literate model called Kosmos-2.5 allows users to upload images, audio, and videos by various methods.

Meta AI’s Long-Context LLMs: Redefining the Landscape of Natural Language Processing - Meta AI's research team has developed long-context Large Language Models (LLMs) that outperform existing open-source models in terms of performance, utilizing continuous pretraining from LLAMA 2 and a modification to positional encoding to handle longer contexts effectively.

Google DeepMind AI tool assesses DNA mutations for harm potential - Scientists at Google DeepMind have developed an AI program called AlphaMissense that can predict whether genetic mutations are likely to cause disease, potentially speeding up research and diagnosis of rare disorders.

Researcher finds way to get audio from still images and silent videos - A researcher has developed a machine learning tool called Side Eye that can extract audio from still images and muted videos by analyzing the tiny bends in light caused by the vibrations of sound, potentially opening up new possibilities for cybersecurity threats and digital evidence in legal cases.

Can robotics take off like GenAI? Moravec's paradox vs. scaling laws - The progress in robotics research, heavily impacted by machine learning, is continuing, but there are still challenges in figuring out the right form factors to sell robots as products, and Moravec's paradox suggests that sensorimotor control is harder to engineer than reasoning and thinking.

Baichuan 2: Open Large-scale Language Models - Baichuan 2 introduces open large-scale language models for individuals and organizations, emphasizing values of openness, community, excellence, and user data privacy.

Meta's new AI chatbot trained on public Facebook and Instagram posts - Meta Platforms used public Facebook and Instagram posts, excluding private posts, to train its new Meta AI virtual assistant, while also taking steps to filter private details from public datasets used for training in order to respect consumers' privacy.

BTLM-3B-8K: 7B Parameter Performance in a 3B Parameter Model - BTLM-3B-8K is a new state-of-the-art 3 billion parameter language model that outperforms existing models and is even competitive with some 7 billion parameter models, providing excellent long context performance while requiring less memory and compute power.

Scientists Say New AI Can Translate What Chickens Are Saying - A team of researchers in Japan claims to have developed an AI system capable of interpreting the emotional states of chickens based on their clucking sounds, potentially leading to a better understanding of the needs of poultry and other animals.

AnyMAL: An Efficient and Scalable Any-Modality Augmented Language Model - An efficient and scalable language model called AnyMAL allows for the augmentation of any modality, including images, audio, and videos, by simply uploading or pasting them into the text input.

Language Modeling Is Compression - The article discusses issues with the paper's argument about PNG as a specialized codec, the lack of clarity on how the picture was encoded, and the misleading impression about FLAC's specialization in storage-efficient lossless audio.

Analysis

The AI assistant revolution is more than 50 years in the making - The article explores the history and potential of AI assistants, discussing their evolution from the 1970s to the present day and the challenges and possibilities they present in terms of natural language understanding and human-like interaction.

Perhaps AI Is Modern Alchemy. And That’s Not a Bad Thing - AI is compared to alchemy, with the author arguing that while AI may lack a robust theoretical basis and rely on trial and error, it has proven to be useful and has the potential to become a new science.

The Fundamental Extrapolation Error - When discussing AI, people tend to make the Fundamental Extrapolation Error by assuming that modern AI is on a path to intelligence like ours, even though it may not be, leading to misguided extrapolations and predictions about its capabilities and risks.

Weird A.I. Yankovic, a cursed deep dive into the world of voice cloning - The article explores the world of AI voice cloning and its implications for creating cover songs, discussing the challenges and limitations of replacing vocals with AI-generated voices and the ethical and legal concerns surrounding the use of these tools.

AI-Generated 'Subliminal Messages' Are Going Viral. Here's What's Really Going On - A new technique called ControlNet allows users to create optical illusions by specifying additional inputs, leading to the viral spread of AI-generated images with "subliminal" messages, although the messages are typically not very subtle and the method for creating them is fairly simple.

Confessions of a Viral AI Writer - A journalist explores the capabilities and implications of AI-generated writing, reflecting on the emotional thread that AI companies might pull on when selling these technologies.

Policy

UK competition regulator lays out AI principles - UK competition regulator proposes principles to prevent AI dominance by tech companies, emphasizing accountability and transparency, and seeks input from major AI developers like Google, Meta, OpenAI, Microsoft.

EU examines Nvidia-dominated AI chip market's abuses, Bloomberg reports - The European Union is investigating alleged anticompetitive practices in the AI chip market, particularly focusing on Nvidia's dominance, according to Bloomberg News.

California governor vetoes bill banning robotrucks without safety drivers - California Governor Gavin Newsom has vetoed a bill that would have banned heavy-duty driverless trucks in the state, providing relief for companies developing autonomous technology for hauling goods.

Indian actor Anil Kapoor wins court battle over AI use of his likeness - Indian actor Anil Kapoor wins court battle over unauthorized AI use of his likeness, setting a progressive precedent for actors in protecting their rights in the face of evolving AI technology.

IBM Tries to Ease Customers’ Qualms About Using Generative A.I. - IBM is addressing concerns about the use of generative AI by offering indemnification against intellectual property claims and publishing its data sets, aiming to ease customers' worries about confidentiality and accuracy.

Google adds a switch for publishers to opt out of becoming AI training data - Google has introduced a new tool called Google-Extended that allows website publishers to opt out of having their data used to train the company's AI models while still being accessible through Google Search.

Amazon restricts authors from self-publishing more than three books a day after AI concerns - Amazon has implemented a new rule limiting authors to self-publishing only three books a day on its site due to concerns about an influx of suspected AI-generated material.

Musicians are eyeing a legal shortcut to fight AI voice clones - Musicians are considering using likeness laws as a legal avenue to protect against the unauthorized use of their voices by AI platforms, sidestepping complicated copyright issues.

AI Security Center to Open at National Security Agency - The National Security Agency is establishing an AI Security Center to oversee the integration of artificial intelligence capabilities into U.S. national security systems.

Fun

Trolling chatbots with made-up memes - Chatbots like ChatGPT, Bard, and GPT-4 can generate realistic but completely fabricated explanations for nonexistent memes, highlighting the problem of retrieving information that doesn't actually exist.

---

Copyright © 2023 Skynet Today, All rights reserved.