Artificial Intelligence - Page 55

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".

An illustrator has taken to X, formerly called Twitter, to accuse Disney of using an AI-generated image as promotional material for its upcoming release of Loki season 2.

The illustrator is Katria Raden, who took to her personal X account with a post that breaks down the image recently published on the official Loki X account. The illustrator explains that the background image of the spiraling clock shows "numerals on the clock turning into meaningless squiggles or just being meaningless squiggles," which are "telltale signs" of an AI being involved in its creation.

Raden went on to publish a link to the stock image that was purportedly used to create the background image used in the Loki promotional. Additionally, the stock image was sourced from Shutterstock and was not labeled as AI-generated. However, content by the same author also contained these "telltale signs" of AI-generated images. Furthermore, Shutterstock policy prohibits any AI-generated content that isn't made with its own in-house AI Image Generator tool.

A new report from Reuters suggests that OpenAI are considering developing its own in-house chips to power the future of AI-tool queries.

The creators behind the immensely popular ChatGPT are reportedly tossing up between acquiring a company to supply GPUs or making one in-house. The report states that OpenAI CEO Sam Altman has prioritized acquiring more AI chips, which will improve the speed and reliability of the API. Additionally, acquiring a new chip source may dramatically reduce running costs for OpenAI, as an analysis by Stacy Rasgon from Bernstein Research suggests that each query on ChatGPT costs OpenAI 4 cents.

Furthermore, ChatGPT received more than 100 million monthly users in its first two months of launching, which equates to millions of queries per day. If OpenAI's chatbot reaches a tenth of Google's total queries per day it would cost the company $48.1 billion in GPUs and $16 billion a year on chips for future years. At the moment NVIDIA is dominating the AI chip-making market, with OpenAI using approximately 10,000 NVIDIA GPUs to power its AI tools.

With the emergence of AI-powered tools used to generate uncomfortably accurate video and audio representations of public figures, we are starting to see a rise in the number of phony advertisements.

AI-powered tools capable of deep-faking actors into promoting specific products have begun to rise, with prominent figures such as Joe Rogan and Dr. Andrew Huberman already having to battle individuals creating ads to sell specific products. Now, Tom Hanks has been added to the growing list as the actor recently took to his personal Instagram account to warn the public of a seemingly AI-generated ad of him promoting a dental plan.

Hanks wrote, "I have nothing to do with it", with Gizmodo being able to track the owner of the image used in the deep fake back to Los Angeles. Another AI story on the same day joined Hank's public warning of the ad. Zelda Williams, Robin Williams' daughter, took to her Instagram Story to issue her own warning about AI being used to create versions of actors without their permission.

In a candid interview, Kevin Scott, Microsoft's chief technology officer, spoke at Vox Media's Code Conference, where he opened up about the recent rise and demand for GPUs for building new hardware and equipping data centers for AI. In addition to noting that NVIDIA's GPU line-up is currently the best fit to deliver AI processing power, he added that supply issues and shortages are improving.

"Demand was far exceeding the supply of GPU capacity that the whole ecosystem could produce," Kevin Scott said during a panel discussion. "That is resolving. It's still tight, but it's getting better every week, and we've got more good news ahead of us than bad on that front, which is great."

With Microsoft investing billions in companies like OpenAU alongside Google building its supercomputer facilities, the demand for NVIDIA's cutting-edge GPUs has seen its stock price rise by 190% in 2023 alone. With Microsoft adding weight to NVIDIA's financial reporting stating that GPU supply will steadily increase each quarter, Kevin Scott also had a few words to say about the competition - and Microsoft's ambitions.

Authorities in France raided NVIDIA offices in Paris last week as part of an anticompetitive investigation in the "graphics card sector." The unannounced seizure operation was carried out to gather information and potential evidence. France's Competition Authority disclosed very little about the situation other than that it was part of a broader investigation into the cloud computing sector.

Thanks to innovation foresight and investment in AI for several years, NVIDIA's GPU hardware, like the A100 and H100, are so far ahead of the competition that the company has secured over 80% of the market. This is a massive lead over rivals like Intel and AMD and one that also has the European Commission formally looking into potential unfair and monopolistic practices in the AI space.

With hardware sales going through the roof for the company and orders reportedly backed up through most of 2024 - the French raid and the EU's investigation are all about understanding NVIDIA's role, its pricing strategies, and more to determine if there's anything potentially shady going on.