Artificial Intelligence - Page 57

ChatGPT appears to have misfired in spectacular fashion at times since yesterday, and although the gremlins in the works may now have been smoothed over by OpenAI, an official investigation into what caused the problems hasn't yet been concluded.

So, what did go wrong? Well, as you can see in the above post on X (formerly Twitter), the AI started talking complete nonsense, sprinkling in Spanglish, and generally acting highly erratically. There are quite a few reports of this kind of behavior, too.

This is the sort of thing we've seen from AI models before, but usually close to launch, when the products are only just out of testing, and still being honed.

According to a new report at Bloomberg, Reddit has signed an AI content licensing deal that will see the social media company receive $60 million annually. For a cool $60 million, an unnamed AI company will get access to every Reddit post and thread for AI model training.

The move comes as a way to boost interest in the upcoming Reddit IPO - with the report stating that Reddit informed potential investors of the $60 million contract earlier this year. Citing sources "familiar with the matter" will give an unknown generative AI model access to decades worth of data, images, clips, and conversations.

Reddit is one of the most popular social destinations on the internet, so a deal like this makes sense when generative AI models need as much data as possible to train on. Until now, using data freely available on the internet (like Reddit posts) has been a go-to for many AI models (including OpenAI's GPT-2), so it's only natural to see money changing hands.

With AI technology and tools advancing at a pace that very few can comprehend, especially regarding applications and possibilities, it's an area where safety and responsibility across all levels are paramount.

This is where the U.S. Artificial Intelligence Safety Institute Consortium (AISIC) comes in, a division of the U.S. Department of Commerce and the National Institute of Standards and Technology.

AISIC will focus on the safety and 'trustworthy' development and deployment of AI, creating tools, methodologies, and industry standards. As a pioneer in AI hardware, tools, and software, NVIDIA has announced that it will work with AISIC to foster an age of AI safety.

NVIDIA will host its GPU Technology Conference (GTC) 2024 event on March 18, returning to an in-person event after the virtual world took over during the pandemic.

The company will be hosting a keynote with NVIDIA CEO Jensen Huang to talk about AI, which shouldn't be a surprise by now, with NVIDIA now promoting GTC 2024 as the "#1 AI Conference for Developers". What should we expect at GTC 2024? NVIDIA will most likely unveil its next-gen Blackwell GPU architecture, and its new B100 AI GPU that will absolutely demolish AI benchmark charts.

NVIDIA's exciting GTC events -- which I used to attend year after year and truly, dearly miss, were one of my favorite events of the entire year in tech -- normally focusing on data centers, workstations, and high-end GPUs. Now, the focus is shifting to the world of AI, which NVIDIA absolutely dominates. It has the most AI GPUs in the market, has 90%+ of the AI GPU market share, and it's about to unleash a next-gen B100 AI GPU.

The delivery speed of NVIDIA AI GPU shipment has been "greatly accelerated," according to UBS analysts; previously, it was taking 8-11 months for deliveries, but now that's been reduced to just 3-4 months.

NVIDIA and TSMC (Taiwan Semiconductor Manufacturing Company) have greatly increased production capacities, solving the bottleneck issue of advanced CoWoS packaging technology. NVIDIA can now not only clear out its H100 AI GPU inventories, but it can begin pushing more of its beefed-up H200 AI GPU shipments, and upcoming next-gen B100 "Blackwell" AI GPUs that are coming this year.

Another factor in AI GPU shipments increasing from NVIDIA is that recent US export restrictions into China have freed up production capacity for other orders. UBS analysts also believe that the decreased delivery cycle implies that short-term growth of NVIDIA AI GPUs is "close to its peak," so continued growth may be weak... but NVIDIA is expecting its AI GPU shipments to increase 150% year-over-year. CEO Jensen Huang also said that the AI market could be worth $2 trillion over the next 5 years recently at the World Government Summit in Dubai.

NVIDIA has officially taken the wraps off of its new Eos supercomputer, which is an incredibly powerful new system that's aimed at AI applications. Check it out:

The new NVIDIA Eos supercomputer features 576 x NVIDIA DGX H100 systems; each of the individual DGX H100 systems has 8 x NVIDIA H100 AI GPUs, delivering a grand total of 4608 x NVIDIA H100 AI GPUs inside powering them. NVIDIA is using its in-house Quantum-2 InfiniBand architecture to connect them all, delivering some truly mind-blowing performance.

NVIDIA's powerful new H100-powered Eos AI supercomputer features 18.4 exaflops of FP8 AI performance, providing NVIDIA with a continued edge against the competition regarding AI performance. NVIDIA first revealed its new Eos supercomputer at the Supercomputing 2023 trade show in November 2023, but now it's here and ready to take on AI workloads.

OpenAI has just announced its latest tool -- Sora -- which will generate videos from text prompts, and it has truly incredible videos out before the new tool has been officially released. Check it out:

OpenAI posted on X about Sora, teasing: "Sora can create videos of up to 60 seconds featuring highly detailed scenes, complex camera motion, and multiple characters with vibrant emotions". The new model is named "Sora" after the Japanese word for "sky" and can produce impressive video results up to 60 seconds long just from a text prompt.

OpenAI explained on its website: "We're teaching AI to understand and simulate the physical world in motion, with the goal of training models that help people solve problems that require real-world interaction".

The United States Patent and Trademark Office (USPTO) has pushed out new guidance on inventorship for AI-powered inventions, stating that only humans can hold patents.

The USPTO said that AI systems can help out during the creative process, but only natural persons (human beings) who make significant contributions to the conception of an invention can be named as inventors. This means that AI models can't just pump out patent ideas without considerable human input, stopping patent hoarding by AIs.

An AI model can't be named as an inventor, or even a joint inventor, on a patent. But, an AI helping out in the creation of that invention doesn't mean that the human being working on it can't hold the patent. The USPTO explains: "While AI systems and other non-natural persons cannot be listed as inventors on patent applications or patents, the use of an AI system by a natural person(s) does not preclude a natural person(s) from qualifying as an inventor (or joint inventors) if the natural person(s) significantly contributed to the claimed invention".

OpenAI has just announced it's rolling out a new memory feature into ChatGPT so that the AI chatbot will remember your past conversations, commands, and requests.

The new memory feature being baked into ChatGPT is coming soon and has been a long time coming. One of my major faults with ChatGPT is that it doesn't remember your conversation, forcing you to re-prompt things. It would be fantastic to remember everything, like an artificial intelligence would... not a limited AI model.

OpenAI explained on its website: "We're testing memory with ChatGPT. Remembering things you discuss across all chats saves you from having to repeat information and makes future conversations more helpful. You're in control of ChatGPT's memory. You can explicitly tell it to remember something, ask it what it remembers, and tell it to forget conversationally or through settings. You can also turn it off entirely. We are rolling out to a small portion of ChatGPT free and Plus users this week to learn how useful it is. We will share plans for broader roll out soon".

NVIDIA CEO Jensen Huang made some big statements at the recent World Government Summit recently held in Dubai, where he said AI is the future of economics and that there'll be $2 trillion worth of data centers powering the future of AI.

Huang added: "We are in the beginning of this new era and what's gonna happen is there's a trillion dollars worth of installed base in data centers around the world and over the course of next 4-5 years, we'll have 2 trillion dollars worth of data centers that will be powering software around the world and all of that is going to be accelerated and this architecture for accelerating computing is ideal for this next generation of software called generative AI".

It's not surprising that Jensen believes the data center market will explode into a gigantic $2 trillion market in the next 4-5 years, which will be the driving force behind the future of everything AI. NVIDIA is leading the pack here, where they're not only commanding an estimated 90% of the AI GPU market, but the company has seen its market capitalization explode over the last few months... adding $650 billion to see NVIDIA with a commanding $1.78 trillion market cap at the time of writing.

NVIDIA's going big on AI - in case you hadn't noticed - and the latest brainwave from Team Green is a new 'Chat with RTX' demo app.

This software allows you to leverage the power of your RTX graphics card - only RTX 3000 and 4000 GPUs are supported, mind - to run a local chatbot tapping into a large language model (the choices are Mistral and Llama, open source efforts of course).

Then you can use your preferred option, hook it up to some files on your PC, and query this local generative AI, letting it comb through those files for answers to queries.

NVIDIA CEO Jensen Huang was recently at the World Government Summit in Dubai, where he called for all countries to build their own "sovereign AI."

Huang said: "This is the beginning of a new industrial revolution, is about the production -- not of energy, not of food -- but the production of intelligence, and every country needs to own the production of their own intelligence, which is the reason why there's this idea called"sovereign AI".

He continued: "You own your own data. nobody owns it. Your country owns the data; it codifies your culture, your society's intelligence, your common sense, your history. You own your own data. You, therefore, must take that data, refine that data, and own your own national intelligence. You cannot allow that to be done by other people".

The insatiable demand for AI GPUs and all things AI throughout 2023 has seen HBM memory makers enjoying a gigantic 500% spike in the average selling price of HBM chips. Not bad at all for HBM makers Micron, Samsung, and SK hynix.

HBM chips are one of the most important parts of an AI GPU, with the likes of AMD and NVIDIA both using the bleeding edge of HBM memory on their respective AI GPUs. Market research firm Yole Group has now chimed in, predicting that the HBM supply will grow at a compound annual rate of 45% from 2023 to 2028, with HBM prices expected to remain high "for some time," considering how hard it is to scale up HBM production to keep up with the crazy demand.

Samsung and SK hynix are the dominant players in the HBM memory-making business; between the South Korean companies, they have 90% of the HBM market share, leaving Micron with the scraps. SK hynix and TSMC (Taiwan Semiconductor Manufacturing Company) are teaming together, so you've got some interesting developments that will continue to roll on as the months (and years) go on.

NVIDIA is arguably the leader in creating GPU hardware for AI, and it shouldn't come as a surprise that it's using AI to accelerate chip design and production. Staying ahead of the game is a part of the picture; NVIDIA is now the world's sixth-largest company thanks to the rise of AI.

NVIDIA's custom LLM (Large Language Model) is called ChipNeMo, adapted from Meta's Llama 2, and is trained on the company's vast amounts of architectural data, documentation, and source code. ChipNeMo was first unveiled in late 2023 and is also used as a tool for training engineers via its chatbot functionality.

According to Bryan Catanzaro, NVIDIA's vice president of applied deep learning research, designing a new GPU takes close to 1,000 people working together. Access to an AI chatbot with all that data is an invaluable tool. The question now becomes, how much - if any - of the design for upcoming GPUs is being designed by ChipNeMo?

Robocallers have been a problem for many years now, with some individuals receiving more calls than others. But with the power of artificial intelligence (AI), the frequency of these calls has increased and seemingly reached its boiling point when an AI-powered robocaller imitated President Joe Biden.

These robocalls are likely familiar to everyone who owns a phone, and with the exponentially increasing power of AI-powered tools, scammers are becoming more and more creative in devising ways to scam individuals out of their hard-earned dollars.

The Federal Communications Commission (FCC) is aware of this issue and was pushed to action when an AI-generated robocall imitating President Joe Biden made headlines as the voice on the end of the line was telling New Hampshire Democrats not to vote for Biden in the upcoming presidential primary.

NVIDIA CEO Jensen Huang has been elected to the National Academy of Engineering (NAE), which is one of the highest academic distinctions awarded to engineers. The NAE credits NVIDIA CEO Jensen Huang "for or high-powered graphics processing units, fueling the artificial intelligence revolution".

It wasn't too long ago that Jensen won the "Best CEO of 2023" by The Economist, but now his "outstanding contributions" to the world of engineering were for both the GPU and AI GPU industries. The NAE takes in a bunch of new members each year after they spend multiple months selecting them.

There are specific qualities, achievements and contributions that potential NAE members are judged by, with the academy only honoring people who have made outstanding contributions in the following fields:

I think by now we all know that the "AI PC" and AI in general is the future of technology, but we haven't been able to run generative AI tasks locally -- as the likes of ChatGPT is run in the cloud -- but that is slowly changing. The entire PC industry is being dragged into AI whether they like it, or not.

AMD and Intel both have processors on the market with NPUs (Neural Processing Units) that are dedicated parts of the chip that are built for AI workloads, but you can also use your CPU or GPU to run generative AI tasks, or a combination of them. CPUs and GPUs aren't exactly optimized for AI workloads, which is where the NPU comes in... the faster it gets, the better AI experiences we'll have on future-gen hardware.

Analyst firm IDC is making some predictions to the AI PC market, now that AI processors are here with NPUs inside of them from AMD and Intel.

Well, I didn't think I'd be writing this story at this stage of the AI game, but here we are: a Russian man has used ChatGPT-4 with a custom AI bot he wrote to scour Tinder for the best date... and he ended up getting engaged to one of the women, all thanks to AI.

Aleksandr Zhadan had a bad breakup with his then-girlfriend in 2021, shortly after he started dating women around Moscow and St. Petersburg in his native country of Russia. He said the dates ended in disaster each time, but that all changed in 2022 when he got access to OpenAI's then-new GPT-3 API.

He then developed an AI bot that would automatically swipe right on profiles that featured two pictures, later integrating GPT-3 into his bot and allowing the LLM to chat with women that met his parameters, all without him touching a keyboard or smartphone. Zhadan used the prompt, "You're a guy talking to a girl for the first time. Your task, not immediately but to invite the girl on a date".

Microsoft has announced improvements to Copilot in what's the latest step forward for the AI (no doubt the first of many this year).

In a blog post, the software giant told us that it has made Copilot more streamlined, with a cleaner and slicker look, as well as introducing a carousel of suggested prompts to stoke your curiosity (perhaps) to explore those topics.

What this means is that the interface for Copilot does indeed look tidier and neater, with that carousel occupying center stage in the middle of the page.

Generative AI and autonomous agents will make their way into government-run programs, including the military, so this little experiment with some of the most well-known LLMs is interesting... to say the least. Like the classic 1983 film WarGames, various AI models were pitted against each other in multiple wargame scenarios to see how they'd react and make decisions.

You can read the full results of the experiment in a new paper titled 'Escalation Risks from Language Models in Military and Diplomatic Decision-Making' from several high-profile universities and institutes. Eight different "autonomous nation agents" using the same LLM were put in a wargame scenario - with a separate AI model summarizing the simulated world's outcomes, consequences, and state.

Turn-based tabletop gaming, except to see what would happen if AI was in charge of every military asset, including nuclear weapons. And yes, a few of the LLMs went nuclear and started dropping bombs.