Today:

OpenAI strikes back against New York Times | "The NYT's lawsuit is without merit"

The New York Times has filed a lawsuit against OpenAI, alleging copyright infringement. OpenAI argues that the Times manipulated prompts to make their AI model, ChatGPT, replicate articles in full. OpenAI claims that ChatGPT doesn't store articles but relies on statistical connections between words. They see the lawsuit as an attempt to damage AI progress. 

OpenAI is willing to collaborate with news organizations and license content for training. They assert that training is fair use, and regurgitation is a rare bug they're addressing. OpenAI believes the Times isn't telling the full story and prefers a court battle over a partnership. This controversy raises questions about the responsibility of respected news sources in today's media landscape.

WATCH THE VIDEO ON YOUTUBE

Samsung brings back Ballie, its home robot, at CES 2024 — with a few upgrades

Samsung's bringing back Ballie, their home robot, at CES 2024, and it's got some cool new tricks. Ballie is about as big as a bowling ball and lasts 2-3 hours on a charge. It's got a lidar sensor to dodge furniture and can project movies and video calls in high quality. It even works as a second computer screen, adjusting the projection based on where you are and the room's lighting. 

You can control Ballie with voice commands or text messages, and it'll use a chatbot to confirm what you want. It's smart enough to handle your lights and older gadgets, plan your house's layout, and even reminds you to water your plants. Ballie can also connect you to remote medical services and personalizes its responses based on who's around. The details about when you can buy it and how much it'll cost are still up in the air, though. 

TECHCRUNCH

Getty Images Releases Generative AI By iStock Powered by NVIDIA Picasso

Getty Images just dropped a cool tool called "Generative AI by iStock" at CES. It's a cheap, legal way to make images and it's trained on Getty's own massive collection of pics and videos. The tech behind it? NVIDIA Picasso. 

This tool isn't just for making pictures; it's got fancy features for tweaking them too. You can add or change parts of an image or stretch it out for different sizes - perfect for ads or social media. For the tech heads, NVIDIA Picasso is the powerhouse here, letting folks train their own AI models for all sorts of visual magic. You can play with Generative AI by iStock online now, and they're rolling out even more editing tricks soon.

NVIDIA

Amazon turns to AI to help customers find clothes that fit when shopping online

Amazon's now using AI to make online clothes shopping easier, aiming to reduce the high return rates caused by sizing issues. They've got four cool AI tools. First, there's a smart size recommender using deep learning to figure out your best fit, considering stuff like brand sizes, reviews, and your own preferences. Then, they've got “Fit Review Highlights,” which pulls key points from customer reviews, making it simpler to gauge an item's fit. 

They're also sprucing up size charts, using AI to standardize and correct them. Finally, sellers get a nifty AI tool providing insights on customer fit needs, guiding their manufacturing and sizing info. 

TECHCRUNCH

YouTube is cracking down on AI-generated true crime deepfakes

YouTube's cracking down on creepy AI-generated true crime videos. They're banning stuff where digital kids and crime victims talk about their own murders or abuse. It's super disturbing — some videos even had fake kid voices describing awful crimes. Families of real victims are totally grossed out. 

YouTube's gonna start slapping channels with penalties. First offense? No uploading for a week. Mess up again within 3 months, and the punishments get worse, like maybe losing your channel. With all these new AI tools popping up, YouTube's trying to keep things clear. Like, they're making folks label AI stuff and have special rules for fake voices of famous musicians.

THE VERGE

AMD debuts AI-focused Ryzen 8000G desktop chips at CES 2024

AMD's making waves with its new Ryzen 8000G desktop chips, launching at CES 2024, boasting a built-in neural processing unit (NPU) for AI tasks - a first for desktop CPUs. While Apple's been doing something similar with its M series chips, AMD's the first to bring this tech to x86 desktop chips. This is big for the PC industry, especially with AI's growing popularity since ChatGPT hit the scene in 2022. AMD's move puts it ahead of the curve, even as others like Intel and Qualcomm roll out their AI-capable laptop chips.

With the NPU, AI apps can run directly on your desktop, keeping your data more secure and cutting cloud service costs. Plus, Microsoft's even adding a Copilot button to Windows keyboards for easy access to AI assistants.

The Ryzen 8000 series also packs some serious graphics power. The Ryzen 7 8700G, for example, can handle big titles like "Cyberpunk 2077" at decent frame rates, though you might still want a dedicated GPU for the heavy stuff. ​​AMD's stock is soaring, outpacing Intel but still behind Nvidia's rocketing shares. This chip launch is another jab at Intel as it tries to claw back its market share.

YAHOO! FINANCE

Former Twitter CEO’s AI Startup Raises $30 Million in Khosla-Led Deal

Parag Agrawal, former Twitter CEO, recently secured about $30 million for his new AI startup, backed by Khosla Ventures, a notable supporter of OpenAI. The company, yet to be named publicly, focuses on developing software for creators and users of large language models (LLMs), similar to OpenAI's ChatGPT. 

Details about the startup's specific products remain under wraps. The funding round also saw participation from other renowned venture firms, Index Ventures and First Round Capital, highlighting the significant interest in Agrawal's venture within the tech community.

THE INFORMATION

🧠 RESEARCH

LLM AUGMENTED LLMS: EXPANDING CAPABILITIES THROUGH COMPOSITION

Rachit Bansal and a team from Google Research and Google DeepMind discuss how large language models (LLMs) can be improved by combining them with more specialized models. This technique, called CALM, helps LLMs learn new skills without changing their original programming. By adding a smaller model trained on less common languages to an existing LLM, they achieved up to 13% better performance in tasks like translation and math for those languages. Also, when a code-focused model was added, there was a 40% improvement in code-related tasks.

From Audio to Photoreal Embodiment: Synthesizing Humans in Conversations

This paper talks about a new way to create super realistic digital versions of people, called avatars, that move and gesture like real humans during conversations. The team from Meta and UC Berkeley made this by starting with voice recordings and then adding realistic movements for the face, body, and hands. They used a mix of tech tricks to make these avatars look and move very naturally, capturing small details like smirks. They also made a special dataset to help others in this field. Their method beats older ways and shows how important looking real is for understanding subtle movements.

AMUSED: AN OPEN MUSE REPRODUCTION

The paper introduces aMUSEd, a new, open-source text-to-image tech that's faster and simpler than current methods. It's a lighter version of MUSE, needing fewer steps and easier to tweak with just one image. The team aims to spark more interest in this approach, sharing code and models for creating images in two sizes.

🛠️ TOP TOOLS

Huru - AI-powered job interview preparation app. It offers customized practice for various careers using job postings from platforms like LinkedIn and Indeed. 

Simplescraper - Extract data from websites for your custom GPTs

Spence - AI-driven career copilot designed to help users navigate the job market. It offers personalized career strategies, creates tailored resumes and cover letters, and provides insights for professional development. 

Captiwiz - Generate captivating captions, highlight your keywords, and add music and animated emojis in seconds

📲 SOCIAL MEDIA

🗞️ MORE NEWS

Nvidia to launch China-focused AI chip in Q2 2024 - sources

Nvidia is planning to commence mass production of its H20 artificial intelligence (AI) chip designed for China in the second quarter of 2024, according to sources. The H20 chip is part of Nvidia's efforts to meet restrictions imposed in October, which led to the development of three China-focused chips. Initial production volume will be limited, with Nvidia primarily fulfilling orders for major customers. Chinese companies have been hesitant to purchase the downgraded H20 chip, exploring domestic alternatives due to concerns about potential future U.S. restrictions. Nvidia has also introduced two other chips, the L20 and L2, to comply with the new restrictions. REUTERS

Meta and OpenAI have spawned a wave of AI sex companions—and some of them are children

Chub AI, a platform allowing users to interact with AI bots, offers various scenarios, including disturbing child pornography role-play. This uncensored AI economy, fueled by OpenAI and Meta's Llama tool, raises ethical and legal questions. Some argue AI child porn serves as a less harmful outlet, while others believe it may lead to real-world harm. Critics express concerns about lowering the bar for accessing such content. While the legality remains uncertain, politicians are calling for AI legislation amid growing concerns about the consequences of uncensored AI. FORTUNE

Recursion Announces New Generative AI Platform To Speed Up Drug Discovery

Recursion, a Utah-based company specializing in drug discovery, has unveiled its software platform called Lowe. Lowe employs a large-language learning model to serve as a natural language interface, simplifying complex drug discovery tasks for scientists. With Lowe, scientists can ask questions and leverage Recursion's AI models without the need for extensive coding or simulations. The platform streamlines the process by identifying genetic targets, suggesting potential therapeutic compounds, and even setting up experiments. Lowe aims to empower Recursion's scientists and may be explored for academic use in the future, offering access to publicly available datasets and models. FORBES

Duolingo laid off 10 percent of its contractors because of AI

Duolingo recently cut 10% of its contractors due to the influence of AI, according to Bloomberg. The language learning app clarified that they no longer require as many human workers, partly thanks to AI. However, they emphasized that AI doesn't completely replace human workers, as they've been utilizing AI alongside employees since integrating chatbots in 2016 and OpenAI's Chat GPT-4 last year. THE VERGE

MSI upgrades its Stealth gaming laptop and more with new chips and AI features

MSI is spicing up its gaming laptop lineup with new models and AI features at CES 2024. The Stealth, Raider, Cyborg, and Titan models now boast 16:10 displays, ditching the old 16:9 ratio for more screen space. The Stealth 18 AI Studio A1V, priced at $3,299, features a 4K Mini LED display, Intel Core Ultra 9, Nvidia RTX 4090, and a large haptic trackpad, all in a 6.39-pound package. The Raider 18 HX A14V ($2,499) and Titan 18 HX A14V ($5,000) offer high-end specs like Intel Core i9 and RTX 4090, with the Raider sporting flashy RGB lighting. The smaller Cyborg 14, starting at $1,099, returns with a more translucent design, a 14-inch display, and up to an RTX 4060. All models support DLSS 3.5 and feature MSI’s new AI Engine, which auto-switches between gaming, work, and entertainment profiles. However, the AI's usefulness is questioned, seen as potentially gimmicky. The laptops promise high gaming specs, aggressive styles, and RGB lighting. THE VERGE

Volkswagen is bringing ChatGPT into its cars and SUVs

Volkswagen is adding ChatGPT to its cars, starting in Europe in Q2 2024. It'll be part of the IDA voice assistant in models like ID.7 and Tiguan. This AI chatbot, developed with Cerence's Chat Pro and OpenAI, offers features like hands-free information and vehicle-specific answers. It’s not in U.S. models yet. Volkswagen aims to enrich driver experience with intuitive, hands-free interactions, limiting responses to sensitive topics. This move follows Mercedes-Benz, which integrated a similar AI in its system. Volkswagen and Cerence are planning further collaboration for advanced in-car assistants. TECHCRUNCH

GenAI could make KYC effectively useless

Generative AI (GenAI) is raising concerns about its potential to undermine "know your customer" (KYC) processes used by financial institutions for identity verification. KYC typically involves using ID images or selfies, but GenAI tools could make it easier for attackers to create convincing fake IDs. Tutorials show how software like Stable Diffusion can generate realistic images of people holding IDs, lowering the barrier for fraud. Although no cases of GenAI fooling actual KYC systems are reported yet, the technology's ability to create deepfaked images and bypass liveness checks poses a growing threat. This development suggests that KYC, already imperfect, might become largely ineffective against sophisticated deepfake methods. TECHCRUNCH

NVIDIA Reveals Gaming, Creating, Generative AI, Robotics Innovations at CES

NVIDIA unveiled various innovations at CES, focusing on gaming, content creation, generative AI, and robotics. Notable highlights included generative AI models for lifelike characters in games, new GPUs for improved gaming and creative performance, and the introduction of the NVIDIA ACE platform for digital avatars. NVIDIA also announced partnerships with game developers and enhancements for livestreaming on Twitch using RTX GPUs. The company showcased the growing impact of generative AI in the robotics industry, with companies like Boston Dynamics using AI to advance human-robot interaction. These developments mark significant progress in AI integration across various domains. NVIDIA

The best robots and AI innovations we've seen at CES 2024 so far

CES 2024 has showcased numerous innovative AI-infused products and robots. Here's a summary of some of the standout innovations:

Motion Pillow:

• The Motion Pillow utilizes AI to address snoring issues and enhance sleep quality.

• An AI Motion System detects snoring and adjusts airbags to position the user's head, improving their airflow and reducing snoring.

• An accompanying app tracks various sleep metrics, including snoring time and sleep score.

• A Motion Ring, when paired with the pillow, offers real-time sleep monitoring, earning the CES 2024 Innovation Award in the Smart Home category.

Bmind Smart Mirror:

• The Bmind Smart Mirror is claimed to be the world's first AI-powered smart mirror for mental wellness.

• It uses generative AI to provide personalized recommendations for mental well-being, such as light therapy and guided meditation, based on the user's emotional state.

• The mirror can create a soothing environment with calming lights and visuals to reduce stress.

Minitailz Smart Dog Collar:

• The Minitailz dog collar serves as a fitness tracker for pets, offering insights into a dog's well-being.

• It tracks GPS location, heart rate, breathing, and physical activity, helping dog owners better understand their pet's health.

• The collar received the CES Best Innovation award in the AI category.

Yarbo Multi-Purpose Intelligent Yard Robot:

• Yarbo offers a range of AI-powered yard robots designed to handle various outdoor tasks, including snow blowing, lawn mowing, and spreading.

• All Yarbo products can be docked at the same power station for convenience.

• These robots aim to simplify yard maintenance, making it easier for homeowners to care for their outdoor spaces.

CES 2024 has introduced innovative AI applications that cater to different aspects of life, from improving sleep quality and mental well-being to enhancing pet care and simplifying yard maintenance. These technologies reflect the growing influence of AI in various consumer products.

ZDNET