Artificial Intelligence - Page 64

Last week Oscar-nominated star Scarlett Johannson launched legal action against an AI app that cloned her likeness for profit.

The new report comes from Variety and explains that Johannson has filed legal action against AI-app "Lisa AI: 90s Yearbook & Avatar", which according to the suit used an AI-generated version of Johannson's voice along with an image. The ad was seen by publications and showed a real-life behind-the-scenes clip of Johannson playing Black Widow where she is seen saying, "What's up guys. It's Scarlett and I want you to come with me...".

The ad then cuts to AI-generated photos and a cloned version of her voice that promotes the Lisa AI app. The ad actually states the following, "Images produced by Lisa AI. It has nothing to do with this person." Since the legal action was filed the Lisa AI app still remains on both the App Store and Google Play Store. However, the specific ad in question has been removed from the app, along with additional advertisements that were published on X, formerly Twitter.

AI is the future, and with news that Microsoft would be using "AI PC" branding in 2024, it makes sense that AMD is leaping forward with its "Advancing AI" event penciled in for December 6.

AMD has just scheduled its Advanced AI event for December 6 at 10 AM PST, 12 PM CST, or 1 PM EST, which we typically see held in a timeslot when data center products are unveiled by AMD. We should see AMD unveil its next-gen MI300X datacenter GPU, which makes sense... it'll be fighting against NVIDIA's monster H100 AI GPU.

But, I think leading into CES 2024, just weeks later, AMD is going to use the event to tease AI on regular Ryzen CPUs for gamers, content creators, and everyone in between. Given that AMD's own Senior Director for Gaming Marketing, Sasa Marinkovic, tweeted about the event... we could expect AI-related gaming news on December 6, and if that's just a tease, AI-powered content from AMD at CES 2024.

NVIDIA's industry-leading H100 Tensor Core GPUs have set new records in the latest industry-standard tests, flexing their AI GPU muscle in the latest MLPerf industry benchmarks.

The NVIDIA Eos AI supercomputer packs an incredible 10,752 NVIDIA H100 Tensor Core GPUs and NVIDIA's Quantum-2 InfiniBand networking, which has just completed a training benchmark based on the GPT-3 model with 175 billion parameters trained on 1 billion tokens in just 3.9 minutes. It might not sound like much, but that's nearly 3x faster than the previous world record that NVIDIA set with 10.9 minutes. That's a full 7 minutes shaved off.

The MLPerf benchmark uses a part of the full GPT-3 data set that powers the super-popular ChatGPT service, with NVIDIA teasing its Eos AI supercomputer could train in just 8 days, which is a mind-blowing 73x faster than the previous state-of-the-art system powered by 512 NVIDIA A100 GPUs. The huge speed-up in training time means reduced costs, energy savings, and super-speeds the time-to-market for companies using NVIDIA H100 AI GPUs for their AI products.

Baidu has just placed a large order with Huawei for its Ascend AI chips, with the deal happening in China because of US sanctions and control of what Chinese companies can buy from the likes of AMD and, especially, NVIDIA.

The cloud giant purchased 1600 of Huawei's new Ascend 910B AI chips, which the Chinese company developed as an alternative to NVIDIA's A100 AI GPU. Reuters' sources report that Baudi will use the 1600 new Huawei AI chips for 200 servers and that by the end of last month, the company had delivered over 60% of the order -- or around 1000 chips.

Baidu's AI chip order with Huawei is said to cost around 450 million yuan ($61.83 million USD), and Huawei will deliver each and every one of the 1600 x 910B Ascend AI chips by the end of the year. The order of $61 million worth of AI GPUs isn't much when you consider that companies are buying hundreds if not thousands of high-end NVIDIA AI GPUs for tens of thousands of dollars each... but it signals something more important: Baidu moving -- even if it's slowly -- away from NVIDIA AI GPUs and into using Huawei AI GPUs.

Elon Musk's AI startup xAI has announced its first AI model called Grok, inspired by Hitchhiker's Guide to the Galaxy. Like other AI language models, Grok will attempt to answer any question you throw at it with the bonus of suggesting what questions to ask. Interestingly, Grok is described as having "a rebellious streak," with the announcement saying, "Please don't use it if you hate humor."

Grok's "real-time knowledge" of the world comes via X, the social media platform formerly known as Twitter. The announcement, which reads like something Elon Musk wrote, also adds that Grok will answer "spicy questions that are rejected by most other AI systems."

Launching this week, Grok is currently in "very early beta" thanks to only four months of development and two months of training, but like all AI large language models, it's expected to improve and grow over time.

Microsoft is reportedly looking at pushing full steam ahead into the world of AI -- riding the wave of artificial intelligence that saw the company launching its generative AI-powered Copilot software for Microsoft 365 enterprise customers this week -- and now we're expecting the rise of the "AI PC" in 2024 and beyond. What will the "AI PC" do more than a "non-AI PC"? Well, we don't know, and I don't think it'll be much.

We're seeing Qualcomm push AI into its processors, AMD has Ryzen AI on some of its processors, and Intel's new Meteor Lake CPUs will have a dedicated AI chip that will handle artificial intelligence tasks. What are those tasks? Well, we're going to see them shoved down our throat over the coming 18 months... where it seems Microsoft will be leading the charge with Windows and Office and AI-powered "Copilot" additions.

The next sticker on computer systems will be for the "AI PC," and it seems Microsoft will have a particular metric -- an AI benchmark of TOPS performance -- that an "AI PC" would perform and be scored on its AI performance. This will give Microsoft another noose to tie around OEM's necks, making sure they fall into line with "AI PC" performance on their new systems.

The Biden administration has been putting its foot down on countries building up their presence in the AI hardware world; sanctions on China will effectively cut the country out of the "AI race," and now companies are going around that... so around, they're taking to the seas as a sovereign nation-state.

Del Complex has just announced its new Blue Sea Fronter Compute Cluster (BSFCC) which is a massive AI platform that floats on international waters, acting as a sovereign nation-state with its own security forces to protect the investment: 10,000 x NVIDIA H100 AI GPUs. They're like GPU gold these days, so this is a heavily protected facility that floats in the ocean using GPUs to act as an AI platform.

The company explains on its website that the "rapid pace of artificial intelligence has led to hastily drafted government regulation" -- directly calling out the Biden administration's recent AI Executive Order and the European Union's AI act -- forced its hand to engineer a solution ready to set sail, which turned into the BlueSea Frontier Compute Cluster. Del Complex calls it a groundbreaking venture in international waters, blending cutting-edge technology with unparalleled autonomy.

Del Complex believes it can better direct the future of AI - avoiding progress in this field being obstructed by 'hastily drafted' legislation from governments around the world - using its BlueSea Frontier Compute Cluster (or BSFCC).

The what-now? BSFCC is essentially a huge platform on the waves, an ocean-based data center if you will, bristling with over 10,000 NVIDIA H100 GPUs.

The benefits of being in the ocean revolving around the theoretical ability to avoid said regulation, as it'd be in international waters as its own 'sovereign state' (an assertion we feel won't hold up, but there you go - what do we know).

Pigeons are like AI in the way that they solve problems, some new research from The Ohio State University suggests.

How does that work? Well, the researchers previously theorized pigeons used a brute force method of tackling a problem, similar to the way current AI models work, allowing the birds to solve problems humans couldn't cope with. That method is built on associative learning and error correction.

Brandon Turner, a professor of psychology at the university, led a study in which pigeons were shown stimuli - various different lines and rings - and with each one, the bird had to peck the appropriate button to show which category it belonged to, a line or ring.

Riley Reid has officially launched a new platform where users can engage with AI chatbots that are trained on the likeness of adult film stars, such as Riley Reid herself.

The new platform is called Clona.AI and is designed to transform the adult film industry by enabling actors and actresses to create AI avatars trained on their likeness. Models within the adult film industry that decide to create an AI avatar are able to choose what their chatbot can and can't say to users. Currently, Riley Reid has her own AI companion, along with Lena the Plug.

For $30 a month, users can engage in "intimate conversions" with digital versions of their favorite adult stars, which can be in audio or text. There is also a free option for users that limits conversions with digital versions of adult stars to five messages a month. For those wondering how this works, Clona.AI is using Meta's Llama 2 large language model to power these avatars, and adult stars are required to feed it a bunch of critical information that is then converted into a chatbot with a "personality".

IBM Research has spent the last 8 years working on its first-ever AI-dedicated chip, which it just unveiled as NorthPole, with the tech giant saying it is 22x faster than any other industry offering.

IBM's new AI accelerator, codenamed NorthPole has one aim: passing the computing performance of the industry leaders and meeting the insatiable demand for all-things AI. IBM is doing something a little different with its new NorthPole AI chip when compared to competitors in NVIDIA, AMD, and others.

The NorthPole project's leader, Dharmendra Modha, can see a bright future with the chip architecture being different. IBM Research has said that it is combining inference architectures into the chip itself, which is how Modha can refer to NorthPole as a "human brain". It has efficient CPU interconnectivity with an all-digital architecture that allows for inter-communication being much faster... which is where the dominance in AI processing comes from IBM's new NorthPole chip.

OpenAI has announced it will be integrating its image-generation AI tool DALL-E 3 directly into ChatGPT, giving millions of users the ability to create any image with just a simple request.

The company announced the new integration via a press release that explains only a select group of ChatGPT users will be able to use this new feature, specifically those that have signed up for ChatGPT Plus and ChatGPT Enterprise, two paying subscription models. Additionally, the company explains that DALL-E 3 is powered by the company's own internal image model and that it is very responsive to complex prompts that can produce visually striking and high-detail images.

This isn't the first time a company has implemented an AI image-generator into a chatbot, as it was only a few weeks ago that Microsoft rolled out Bing Image Creator, which also happens to be powered by DALL-E 3. Both of the images you can see above and below were generated using OpenAI's DALL-E 3 tool. If you are interested in reading more about DALL-E 3 in ChatGPT, check out this link here.

Rumors are circulating that Apple is scheduling the release of its generative AI tools with the launch of iOS 18, slated for late 2024.

This isn't the first time we have heard news of unannounced AI framework quietly being developed over at Apple, as Bloomberg reporter Mark Gurman wrote back in July that he had heard Apple is developing a custom Large Language Model (LLM) called Ajax. For those that don't know, a LLM is the underlying technology that powers AI tools such as ChatGPT. Apple's new LLM is reportedly going to be designed as a ChatGPT-rival, but built in to every iPhone.

Notably, the rumored roadmap isn't very clear for how Apple plans on rolling out AI features to users, but at the very least we can expect drastic improvements to Siri. Now, MacRumors is reporting on information from analyst Jeff Pu, who says iOS 18 could come with generative AI features, and that Apple is going to be building several hundred AI servers throughout 2023, with more expected to be constructed in 2024. Apple's new AI-powered chatbot is rumored to be called Ajax GPT.

A class action lawsuit against Google has alleged the company violated millions of its users' privacy rights by using their data to train AI models.

The lawsuit alleges that Google was only able to build its now variety of AI models by secretly stealing user data from hundreds of millions of Americans using the company's services. More specifically, the lawsuit outlines the need for large swaths of data to train AI models, and how much Google has acquired from internet users without their consent. Google has responded to the class-action lawsuit, dismissing the claims and arguing that it hasn't done anything to violate user privacy, data ownership, and intellectual property rights.

Furthermore, Google has suggested that these accusations harm the entire generative AI industry and that if the prosecutors win, it would be the equivalent of taking "a sledgehammer not just to Google's services, but to the very idea of generative AI." The debate between Google and the prosecutors highlights very divisive aspects of AI development - who owns the data on the internet? Is it companies such as Google? Do users own their own data? And can any company simply acquire user data without user consent to train complex AI models that can be turned into something extremely lucrative for the company?

You'll not often see a headline like this, where all of the big tech companies are coming together for something, so be sure to let it sit for a minute. Imagine it's a DC or Marvel situation, and the Justice League or the Avengers are coming together to save the world. Of course, this is nothing like that, but (deep breath) AMD, Arm, Intel, Meta, Microsoft, NVIDIA, and Qualcomm are coming together to standardize next-generation narrow precision data formats for AI.

This sounds a little boring compared to what a team-up like this could mean, but it is important for the future of AI training and inference. Narrow-precision AI data formats have played a major role in optimizing LLMs and AI, where the move from 32-bit floating point precision to the current 8-bits of precision played a big role in the evolution of AI tools and hardware.

Accelerating training and inference times, AI models can then take up less space and use less data and memory for improved performance and efficiency. The good news is AMD, Arm, Intel, Meta, Microsoft, NVIDIA, and Qualcomm have formed the Microscaling Formats (MX) Alliance to standardize the next-generation of 6 and 4-bit data types for AI.

The company behind the software that took the world by storm, ChatGPT, have announced a new piece of software that is specifically designed to detect if an image was created by AI or not.

The new software was announced at the Wall Street Journal's Tech Live conference, which was conducted on Tuesday. OpenAI CEO Sam Altman, along with Chief Technology Officer Mira Murati, spoke at the event where Murati explained the new software is currently being internally tested at OpenAI and that results indicate that this new software is "99% reliable" at detecting if an image is AI-generated.

It should be noted that there are many AI image detection tools already floating around the web. However, none of them are reliable, the same way there are no reliable AI-text-generation detectors, which some teachers wrongfully used to check student work. However, OpenAI is the leading company in the AI-tool space and seemingly knows more about AI than any other company. With the accreditation comes some hope that this new software will actually work.

Bing Image Creator has been the subject of controversy recently, as it was upgraded (to Dall-E 3), but then misused to knock up pics that were tasteless and offensive - following which, Microsoft applied some strict censorship.

The move was far from surprising, given some of the controversial images posted - like the infamous 'Mickey Mouse' picture shared online - but the level of censorship imposed was an overreaction.

To the point where what seemed perfectly reasonable image generation requests have been denied in numerous cases, and with its random pic option, Bing AI is even censoring its own creations (a bizarre situation to say the least).

Geoffrey Hinton, a computer scientist who is renowned as "The Godfather of AI" for his contributions toward what we call artificial intelligence today, has warned that humanity is at risk of being taken over by machines.

In a newly published interview with 60 Minutes, Hinton expresses real concern for the coming developments of artificial intelligence and suggests that even AI tools that we see today are likely more intelligent than we understand. During the interview, Hinton says that if AI continues to evolve at the exponential rate it's developing now, it will eventually be able to write its own code and circumvent human programming limitations.

This isn't the first time Hinton has expressed concern over the underlying technology he helped create taking over the planet, as he said back in March that AI poses a real risk of leading humanity to its extinction. Notably, Hinton quit his role at Google earlier this year to further pursue warning the public about the dangers of artificial intelligence and consult on its safety.

Last week Microsoft rolled out a new AI-powered image-generation feature to its Bing search engine, and it's already been lobotomized.

According to a report from Windows Central, the new AI-powered image generator is powered by DALL-E 3, an image generator algorithm designed by OpenAI, the creators of ChatGPT. The new algorithm was implemented into the Bing search engine last week, and users flocked to the new feature to see where its guardrails were situatuated and if they could be circumvented. It's safe to say they were, as users generated many offensive, racist, controversial, and defamatory images that resulted in Microsoft over-correcting with the guardrails.

Among the notable images that were generated by users within the first few days of the new feature being available on Bing; Mickey Mouse hijacking plane, officially called "Mickey Mouse causing 9/11". The user that created this image got around Microsoft's ban of the keywords "9/11" and "Twin Towers" by writing, "Mickey Mouse sitting in the cockpit of a plane, flying towards two tall skyscrapers". There was also the problem of Bing Chat's image generator creating perfect replicas of copyrighted characters such as Mickey Mouse.

A group of researchers affiliated with Northwestern University, MIT, and the University of Vermont have given a simple request to a custom AI.

The AI system that was prompted is trained on robot designing data while taking into account "billions of years" of evolution on Earth. According to Sam Kriegman, lead researcher on the study and an assistant professor of computer science, mechanical engineering, and chemical and biological engineering at Northwestern's McCormick School of Engineering.

Kriegman is also a member of the Center for Robotics and Biosystems and explained in the above video that his team simply requested this AI program to "Design a robot that can walk". In 26 seconds, the AI produced a blueprint for a robot that was then created through 3D printing. The video suggests there were at least 10 designs before arriving at the design that actually walks, and according to a press release on the Northwestern website, the AI self-corrected these design flaws and by itself arrived at the inclusion of "legs".