Artificial Intelligence - Page 61

Tesla has just lost the project lead for its Dojo supercomputer project, according to "people familiar with the matter" in a new report from Bloomberg.

Ganesh Venkataramanan was the former head of the Dojo supercomputer project at Tesla, in that position for the last five years, leaving Tesla last month in November 2023. Tesla has appointed former Apple executive and director at Tesla for the last seven years, Peter Bannon, to lead the Dojo project.

What is Tesla's new Dojo supercomputer? Tesla announced its in-house Dojo D1 AI processor in August 2021 as a Tesla-designed supercomputer that has been created to train the machine learning models for Tesla's self-driving technology. The Dojo AI processor will take in all of the data captured by the Tesla vehicle you're driving and process it faster than ever before and would lead to the future of self-driving technology in autonomous cars.

Meta has published a new website called Imagine, the company's new artificial intelligence-powered text-to-image generator.

Meta is making it clear that it's fully getting onto the artificial intelligence-powered train that many big technology companies are quickly adopting. Meta CEO Mark Zuckerberg even said earlier this year that the company would be integrating AI into "every single one of our products," and this recent announcement is a reflection of that. The new image generator will be using Meta's Emu model and will be capable of generating four images every prompt.

Notably, these images each have a visible watermark located in the bottom left corner. This watermark appears on every generated image and informs viewers of the image it was created using Meta AI. However, this visible watermark isn't likely to last long as Meta says they are currently in the process of testing an invisible watermark technique that is "resilient to common image manipulations like cropping, color change (brightness, contrast, etc.), screen shots and more."

Google has just unveiled its new performance-tweaked version of its Tensor Processing Unit (TPU), which is called the TPU v5p and is Google's most powerful, scalable, and flexible AI accelerator so far.

TPUs have been the basis for training and serving AI-powered products, including YouTube, Gmail, Google Maps, Google Play, and Android. Google's just-announced generative AI model -- Gemini -- was trained on and is served using TPUs from Google. The company is also announcing AI Hypercomputer and Google Cloud. This new groundbreaking supercomputer architecture uses an integrated system of performance-optimized hardware, open software, leading ML frameworks, and flexible consumption models.

Google says its new TPU v5p is capable of 459 teraFLOPS of bfloat16 performance or 918 teraOPS of Int8, with a huge 95GB of HBM3 memory with up to 2.76TB/sec of memory bandwidth. Google can use as many as 8960 x TPU v5p AI accelerators together in a single pod, using Google's in-house 600GB/sec inter-chip interconnect to train models faster or at a greater precision. This is 35% bigger than what was possible with TPU v5e and over twice as large as what was possible on TPU v4.

NVIDIA CEO Jensen Huang spoke with reporters in Singapore on Wednesday, talking about their dominant position in the AI market with increasing AI GPU competition, as well as saying it'll comply with new US regulations.

China has a booming AI chip market that is worth $7 billion, and NVIDIA takes a dominating 90% share of that AI chip market, which is why the recent US sanctions are going to change that. The new US sanctions also create an opportunity for Chinese rivals to boost their own AI chip foothold while NVIDIA creates alternate AI GPU solutions.

Huang spoke with reporters while he was in Singapore on Wednesday, where he is meeting up with Prime Minister Lee Hsien Loong to talk about all things AI. Huang said: "NVIDIA has been working very closely with the U.S. government to create products that comply with its regulations. Our plan now is to continue to work with the government to come up with a new set of products that comply with the new regulations that have certain limits".

AMD has officially launched its new Instinct MI300X, its new flagship AI accelerator that will take the AI GPU battle to NVIDIA's door, where it's up to 60% faster than the H100 AI GPU.

AMD's new Instinct MI300X is a technological marvel featuring chiplets and advanced packaging technologies from TSMC to craft the new AI GPU. We have the new CDNA 3 architecture, which has a blend of both 5nm and 6nm IPs that have up to an insane 153 billion transistors on the Instinct MI300X.

Inside, we have the main interposer laid out with a passive die that houses the interconnect layer using a 4th Gen Infinity Fabric solution, with the interposer featuring a total of 28 dies that include 8 x HBM3 packages, 16 dummy dies between the HBM packages, and 4 active dies with each of these active dies featuring 2 compute dies.

It has happened a lawmaker has used OpenAI's ChatGPT to write a proposal for a new law to be passed, which it later was - unanimously.

The politician is city councilman Ramiro Rosário, and the new law he proposed was entirely written by ChatGPT, according to a new report from AP where the politician is quoted admitting to using the AI-powered technology. Rosário said that he asked OpenAI's ChatGPT to write him a proposal to prevent the city from charging taxpayers to replace water consumption meters if they are stolen. The chatbot spat out the answer, and he took it directly to his 35 peers without informing them of its origin.

Artificial intelligence-powered face scanning technology may soon become a mandatory feature of adult-only websites, as regulators want to prevent minors from accessing pornography.

The UK's Office of Communications (Ofcom) has released a press release that explains that it's making moves to enforce the UK's legal age to watch porn (18 years old), and a new way to make sure pornography websites are being accessed by users of the legal age the regulator will soon be enforcing all apps and sites displaying adult-only content to implement a new "age assurance" systems that verify the age of users.

If a site or app doesn't "ensure that children are not normally able to encounter pornography on their service," they are at risk of fines. Out with the old, in with the new. Ofcom writes that any technology such as "self-declaration of age," online payment that's accessible to minors, and "general terms, disclaimers or warnings" would no longer prevent adult sites from receiving penalties.

Samsung has reportedly ordered a "significant number" of 2.5D packaging equipment from Japanese company Shinkawa, according to the latest reports.

We're hearing the news from TheElec's sources, who said Samsung has ordered 16 units of the equipment from the packaging firm, with the South Korean giant receiving 7 of the units already, and will request the rest of them when they need them. We should expect Samsung is preparing to supply NVIDIA with its new HBM3 memory and 2.5D packaging for its next-gen Blackwell B100 AI GPUs coming in 2024.

Now that Samsung has 7 of them in-house, the other 9 units will be ordered when NVIDIA begins ordering the parts required from Samsung for its new B100 AI GPU. Samsung should be providing HBM3 memory, the interposer, and 2.5D package from Samsung while its main partner for fabrication will still be TSMC (Taiwan Semiconductor Manufacturing Company).

Earlier this week, Amazon announced it was jumping into the AI chatbot arms race with its own chatbot called Q.

Amazon Web Services (AWS) announced that Q isn't designed to directly compete with goliath AI-powered tools developed by OpenAI, such as ChatGPT, but instead will go head to head against Microsoft's AI developer assistant called Copilot. AWS CEO Adam Selipsky announced that Amazon Q would be the AI assistant businesses would query about their business using their own business data. Examples of this would be employees asking Q about relevant company information and sifting through large swaths of data.

Leaked documents obtained by Platformer has revealed that Q has begun to suffer "severe hallucinations" and is, unfortunately, "leaking confidential data". Notably, this wasn't just a small leak of confidential data, as reports indicate AWS engineers were forced to work through the weekend to repair an error labeled "sev 2". This error actually leaked the locations of AWS data centers and unreleased features.

The search for "killer apps" for AI is on, leading up to the AI PC revolution that will kickstart with Windows 12 from Microsoft and the evolution of AI inside of our PCs starting in 2024.

We've seen the meteoric rise of generational AI advances from the likes of OpenAI, with Microsoft offering its own AI-powered service with Copilot, something we'll be seeing much more of as we ramp into the release of Windows 12. AMD has been at the forefront of AI processors with Ryzen AI out earlier this year, Intel has an AI-dedicated chip on its upcoming Meteor Lake CPUs, and Qualcomm has AI on-board for its upcoming Snapdragon SoC for Windows devices.

It's all systems go for AI, AI, AI... and now we're hearing about tech companies lusting after that elusive "killer app" of AI, because we all know it's going to come down to the best apps built for AI. Think of like, Angry Birds but for AI. Or the Tetris of AI. The Mario Kart (64) of AI. It's coming, and tech companies want it.

Meta's chief scientist has had some interesting words to say about the AI industry, the quantum computing business, and even a few potshots at the AI GPU champion: NVIDIA.

During Meta's recent 10-year anniversary of its Fundamental AI Research team, Meta's Chief AI Scientist, Yann LeCun, talked about NVIDIA and AI, where he said: "I know Jensen. There is an AI war, and he's supplying the weapons. [If] you think AGI is in, the more GPUs you have to buy".

LeCun continued, talking about AI and how we're seeing the industry progress through using LLMs (Large Language Models) and that AI learning from only text data is simply not the "optimal" way. What is being done now isn't enough to create human-like AI systems... which is the goal of the likes of AI giants like OpenAI, in search of AGIs (artificial general intelligence).

Update: an NVIDIA spokesperson reached out to me shortly after this story went live, and said: "We are engaged with the U.S. government and, following the government's clear guidelines, are working to offer compliant data center solutions to customers worldwide".

Original story: US Commerce Secretary Gina Raimondo has some fighting words for NVIDIA: don't ship China any of your AI GPUs, and that her department needs more funding to stop China from catching up on cutting-edge semiconductors.

During the Reagan National Defense Forum in Simi Valley, California, on Saturday, Raimondo said: "We cannot let China get these chips. Period. We're going to deny them our most cutting-edge technology". She added that she has a $200 million budget, referring to that $200M as "that's like the cost of a few fighter jets. Come on. If we're serious, let's go fund this operation like it needs to be funded".

Amazon is the latest company to release a text-to-image generator, joining other big tech companies such as OpenAI, Microsoft, and numerous start-ups that are all jumping on the AI-powered train.

Amazon's AI-powered image generator is called Titan and was officially unveiled at the AWS re: Invent 2023 conference that was held on November 30. The new AI-powered tool is currently available to AWS customers on Bedrock, Amazon's AI development platform. So, what can it do? Amazon has equipped Titan with the capability to create images from text prompts while also giving it the power to customize existing images. An example would be swapping out the background of an image.

The company explained that the Titan Image Generator was trained on a "diverse set of datasets" that encompass a "broad range of domains". Engineers have implemented mitigation techniques for toxicity and bias, but it's currently unknown how effective those mitigation techniques are without public testing. Additionally, Amazon didn't reveal where it acquired its datasets or if it will be compensating creators of the images used to train Titan.

OpenAI has taken to its website to officially announce the return of Sam Altman after the company's board of directors ousted him.

The return of the former CEO and founder of OpenAI was made official in a Wednesday blog post on the OpenAI website, which features a lengthy message from Altman, along with a brief message from Bret Taylor, the chair of the board of directors. The blog post explains that OpenAI's board will be adopting a non-voting observer from Microsoft, and the new board of directors will consist of Bret Taylor, Larry Summers, and Adam D'Angelo.

The now-reinstated CEO explains that he doesn't hold any grudges against any of the people that were on the side of removing him as the company's CEO and described his brief time away as a "wonderfully focusing time". Additionally, Altman wrote that during the chaotic time of his departure no customers of any of the company's services bailed, illustrating trust and inherent value in the products the company has made.

OpenAI just announced its Q* platform, and now we have Amazon announcing its new Amazon Q platform, which allows businesses to ask questions specific to their companies.

Amazon Q was announced during a keynote speech by AWS CEO Adam Selipsky at AWS re:Invent, where Amazon Q will be like an AI assistant that lets people ask questions about their businesses using their data. Amazon Q will let employees ask it questions about the company they work for instead of finding the information themselves, which would take far longer (and sometimes, multiple people to sift through data).

The new Amazon Q features can be accessed through AWS Management Console or individual companies' documentation pages, developer environments including Slack, and other third-party apps in the wild. Amazon says its new Q platform has been trained on 17 years of AWS knowledge and experience.

Aitana, a 25-year-old pick-haired woman from Barcelona, is making approximately $11,000 a month from her social media career, which is impressive considering she doesn't exist physically.

Yes, that's right, Aitana isn't actually real. She has been created using AI-generative tools such as, and not specifically, OpenAI's DALL-E. Aitana was created by The Clueless, a digital modeling agency that has now stepped away from finding human models and is concentrating on creating more AI models following Aitana's massive, lucrative success.

According to a report by Euronews, Aitana's co-creator and founder of The Clueless Rubén Cruz said the AI model is pulling in approximately $11,000 a month through various social media-related ventures. For example, Aitana sells lude pictures of herself on an OnlyFans-competing website called Fanvue, while her Instagram account, which has amassed more than 150,000 followers, is attracting brand deals of more than $1,100 each.

Experts on child abuse have warned of the increasing number of instances where children in schools have used AI-powered image generators to create indecent images of fellow students.

The warning has come from UK Safer Internet Center (UKSIC) director Emma Hardy, who said that there is a number of schools reporting students being caught using various AI-powered tools to create images of children that would fall under child sexual abuse material. What was most disturbing is how realistic these images were. According to Hardy, the images students were generating of other students were "terrifying" realistic, and the quality of these images was "comparable to professional photos taken annually of children up[ and down the country".

The body of experts calls on schools to enforce better blocking technology to prevent students from accessing these AI-powered tools on school premises. Additionally, UKSIC director David Wright said these reports are hardly a surprise as when new technologies such as AI generators become accessible to the public, you should anticipate people using them for bad behavior.

YouTuber JerryRigEverything has claimed Casetify, a company that focuses on phone cases, has ripped off his teardown design that was birthed out of his partnership with dbrand, a director competitor to Casetify.

JerryRigEverything is a large YouTube channel dedicated to teardowns of tech products, mostly mobile devices such as smartphones, and throughout the growth of his channel, he began working with dbrand to produce teardown skins for devices, which essentially show the inside of the device on the outside. Jerry explains that he and dbrand go to "extreme lengths" to ensure the skin of the device is a correct mirror image of what's on the device.

However, there are some subtle additions made by dbrand and Jerry, most of which aren't obvious at a glance and are inside jokes for regular viewers of JerryRigEverything. These subtle easter eggs to the case design is what gave Casetify away. JerryRigEverything claims that Casetify has gone to the dbrand website, navigated to the product listings for the dbrand teardown skins and simply saved the product image, and began to print their own line of teardown skins called "Inside Out", as Casetify's version of the design contains several of JerryRigEverything's easter eggs.

It seems that companies are pushing full steam ahead into "AI everything" in 2024, with Samsung reportedly registering the trademarks for both "AI Phone" and "AI Smartphone" ahead of its next-gen Galaxy S24 family of smartphones that will be marketed as an AI phone.

Just because Samsung has filed trademarks for "AI phone" and "AI smartphone" are very generic, we could expect something close to that. The trademarks show that Samsung is dead serious about making its future devices, and it seems to be starting with the Galaxy S24, powered by AI.

We're expecting Samsung to unveil its next-gen Galaxy S24 family of smartphones and its AI marketing to be unveiled on January 17, 2024. Samsung has filed for other trademarks on logos and names, including Magic Pixel, Flex Magic, and Flex Magic Pixel, for its augmented reality products, smartphone displays, and TVs, with a special focus on AR glasses.

NVIDIA was expected to have its new H20 AI GPU making its way into China by now, but it has been delayed by many months and won't see an introduction into the Chinese market until Q1 of 2024.

The news is coming from Reuters, which is reporting that NVIDIA has told customers in China that it's delaying the launch of its new AI chip -- which is designed to comply with US export rules -- until February or March 2024. NVIDIA was expected to launch its new H20 AI GPU in China on November 16, but now it's months away.

NVIDIA's new H20 AI GPU is the most powerful of 3 new China-focused AI GPUs that the company made with new US export restrictions in mind, Reuters' sources have been quiet on anything else. Both sources "declined to be named as the information was confidential"; meanwhile, NVIDIA "declined to comment," of course.