#AI Accelerators

[buysellram]

BuySellRam.com is expanding its focus on AI hardware to meet the growing demands of the industry. Specializing in high-performance GPUs, SSDs, and AI accelerators like Nvidia and AMD models, BuySellRam.com offers businesses reliable access to advanced technology while promoting sustainability through the recycling of IT equipment. Read more about how we're supporting AI innovation and reducing e-waste in our latest announcement:

[foxgirltail]

Yes, microsoft trying to make a "zero-water" data center (https://finance.yahoo.com/news/microsoft-unveils-zero-water-data-170002064.html) is unambiguously a good thing. Obviously any reduction of pollution or water usage is a good thing. No, I don't think that means that ai's usage of water is something to singlehandedly be up in arms about.

By all means, be upset about ai! Just don't only be upset about ai for this or that when basically every other industry on the planet has the same exact problems.

A single cotton t-shirt requires 2700 liters of water, 5 trillion liters of water are used annually for fabric dying, and 20% of all water pollution is from garment production (source: https://www.wri.org/insights/apparel-industrys-environmental-impact-6-graphics)

This medium article (https://medium.com/@notkavi/stop-acting-like-ai-uses-a-lot-of-water-fafea5573c63) compares the numbers cited in the same study as the latest news articles about ai water usage ("Making AI Less 'Thirsty' by Pengfei Li et al https://arxiv.org/pdf/2304.03271) to the water needed to make beef. GPT-3 used as much water to train as it takes to feed 12 usamericans their average annual burger-patty supply. One quarter-pounder uses as much water as 36k GPT-3 queries or 3.6k GPT-4 queries.

Here is a comparison of Microsoft's water usage in 2022 to the water usage of golf courses:

In 2022, Microsoft claimed it used 1.7 billion gallons (6.44 billion liters) of water. Between 2003 and 2005, the golfing industry used approximately 2.08 billion gallons (7.87 billion liters) of water DAILY for course irrigation. (water usage of golf courses has dropped an estimated 29% between 2020 and 2005 - https://www.gcsaa.org/media/news-release/2022-news-releases/2022/07/26/golf-courses-reduce-water-usage-by-29-percent-according-to-national-survey - this still amounts to ~1.48 billion gallons or ~6.74 billion liters daily. 2 days of golf is more than a year of Microsoft's water usage)

Source for golf water use: https://www.usga.org/content/dam/usga/pdf/Water%20Resource%20Center/how-much-water-does-golf-use.pdf

Source for ai water use: https://www.standard.co.uk/news/tech/ai-chatgpt-water-power-usage-b1106592.html

Yes, the world would benefit from less water usage for ai. But when you posit ai as uniquely evil for it's water usage, I have a hard time believing you actually have an issue with the water usage. Instead it seems like you just want a reason to dislike ai.

My goal is for all of these industries, and many more, to severely reduce their water consumption, or to even stop existing as industries at all! Is that yours as well? If so, why doesn't it feel like it.

[aipurjopa]

mods make them do stoichiometry

NO NO NO NO I just got HORRIBLE flashbacks I failed stoichiometry so bad when I took chem oh my god 😭 but yes… SUBJECT THEM TO STOICHIOMETRY RN MAKE THEM SUFFER

[howtomakeyousee]

[mittensmorgul]

There’s another post going around about this, but tumblr won’t let me reblog it but...

When I read a story written by a human being, I’m not just reading it because I want to read a coffee shop AU with a specific plot description. I’m reading it because it’s making a connection to another human storyteller and seeing a piece of them carved into the words. Storytelling is a human act of sharing joy, angst, tension, resolution, satisfaction. It’s an act of love.

Writing and reading a story isn’t just an act of creation and consumption. I hate that commercialism and AI are reducing it to that sort of transaction. Like oh, you need words on this subject and that’s the end of it. Like what we really needed was just a vending machine we can push buttons on to get a fix, as if the human creating the story wasn’t a factor. That the author’s life experience and views and feelings haven’t infused the words with their own unique touches.

I’ve read hundreds of coffee shop AU’s over the years (and thousands of fics in general). I’ve seen many similar tropes reused across stories, and just like an AI would, I’ve learned things about writing them that I will always carry with me. But unlike an AI, a human author is not just the sum total of coffee shop AU’s we’ve consumed. Even if we used the same prompt, the same sets of tropes, the same characters. I will always choose the human-crafted story over the computer generated one.

Because again, I’m not just looking for a very specific fix via a series of words. I’m looking for a human connection through story.

Unlike an AI, I have BEEN to a coffee shop. I’ve had experiences in coffee shops. I’ve had funny little meet-cutes with people. I’ve accidentally spilled coffee on myself and knocked heads with someone as we both rushed to wipe it up. I know what it FEELS like. The machine doesn’t.

I’ve also read millions of things that aren’t fanfic, or coffee shop AU’s. I’ve experienced things OTHER than going to coffee shops and having meet-cutes. And I know what all those things feel like when processed through my personal human lens of experience, which is different from every other personal human lens of experience.

All the machine can do is spit out what it THINKS a human experience is, and I honestly don’t care about that at all. Fic is not a “product” to be “generated.” It’s an art form that connects us to other people who share the same love of a thing that we do.

People who, even when all writing the same characters in the same setting to the exact same prompt, will all add something or have a viewpoint about something or bring a completely different personality and life experience to the story that no one else on the planet could. That’s what I’m actually reading.

[fire-on-fuel]

very similarly to how the ai in education argument highlights the issues with academia in the first place I think ai in fandom highlights its issues as well. sure you can use ai for quick roleplaying just like you can use ai for assignments that would be more beneficial for the student to complete without it, but the majority of the stuff under scrutiny is slop.

The observable problem with academia is that the structures of it demand students output as pointlessly and efficiently as an llm, but with fandom it's harder to approach because no one wants to talk about the commodification and consumption of hashed and rehashed and triple hashed content. Even in the example I used about roleplaying, the motivation for that is to get a workable product quickly to skip to the activities in focus.

You can put it in idealist terms and say whoever does that doesnt have the "love for the game" but the reason for the phenomenon is the point of interest existing shifted towards consumption and away from creation. We all joke about tags being flooded by x reader imagines but honestly- there is a demand for slop driven by what informs fandom culture. either the people on the slop only diet make an llm spit it out or they start discourse over why no one's making it for them for free. The role of being a creator is just not in the equation here at all so arguing it with these hyper-passive consumers is pointless.

honestly I think it's reflective of the greater culture shifts that influence western fandom exagerrating. The less niche and perverted (using that word in the social division way) your interests are the closer the fandom experience will be to a sort of clout fueled service industry. this is where you see shit like creators getting frustrated that archive site mechanics like ao3's won't give them a useable clout output like social media sites, senses of entitlement for what is understood by the audience as voluntary but by the artist as unpaid labor requiring a form of payback. the split between monetization and communal collaboration blurs hard.

anyways my main point is that if you like a popular thing, and you like popular fandom cliches within the popular thing, there's no reason to be surprised or reactionary or upset a chatbot can do the same thing as the hundred of creators you look to for regular content: regurgitate in new and interesting patterns. or to be confused when the chatbots and llms are given the same level of reverence as those creators, which is a pedestal of officiality and nothing more

[jcmarchi]

Industry First: UCIe Optical Chiplet Unveiled by Ayar Labs

New Post has been published on https://thedigitalinsider.com/industry-first-ucie-optical-chiplet-unveiled-by-ayar-labs/

Industry First: UCIe Optical Chiplet Unveiled by Ayar Labs

Ayar Labs has unveiled the industry’s first Universal Chiplet Interconnect Express (UCIe) optical interconnect chiplet, designed specifically to maximize AI infrastructure performance and efficiency while reducing latency and power consumption for large-scale AI workloads.

This breakthrough will help address the increasing demands of advanced computing architectures, especially as AI systems continue to scale. By incorporating a UCIe electrical interface, the new chiplet is designed to eliminate data bottlenecks while enabling seamless integration with chips from different vendors, fostering a more accessible and cost-effective ecosystem for adopting advanced optical technologies.

The chiplet, named TeraPHY™, achieves 8 Tbps bandwidth and is powered by Ayar Labs’ 16-wavelength SuperNova™ light source. This optical interconnect technology aims to overcome the limitations of traditional copper interconnects, particularly for data-intensive AI applications.

“Optical interconnects are needed to solve power density challenges in scale-up AI fabrics,” said Mark Wade, CEO of Ayar Labs.

The integration with the UCIe standard is particularly significant as it allows chiplets from different manufacturers to work together seamlessly. This interoperability is critical for the future of chip design, which is increasingly moving toward multi-vendor, modular approaches.

The UCIe Standard: Creating an Open Chiplet Ecosystem

The UCIe Consortium, which developed the standard, aims to build “an open ecosystem of chiplets for on-package innovations.” Their Universal Chiplet Interconnect Express specification addresses industry demands for more customizable, package-level integration by combining high-performance die-to-die interconnect technology with multi-vendor interoperability.

“The advancement of the UCIe standard marks significant progress toward creating more integrated and efficient AI infrastructure thanks to an ecosystem of interoperable chiplets,” said Dr. Debendra Das Sharma, Chair of the UCIe Consortium.

The standard establishes a universal interconnect at the package level, enabling chip designers to mix and match components from different vendors to create more specialized and efficient systems. The UCIe Consortium recently announced its UCIe 2.0 Specification release, indicating the standard’s continued development and refinement.

Industry Support and Implications

The announcement has garnered strong endorsements from major players in the semiconductor and AI industries, all members of the UCIe Consortium.

Mark Papermaster from AMD emphasized the importance of open standards: “The robust, open and vendor neutral chiplet ecosystem provided by UCIe is critical to meeting the challenge of scaling networking solutions to deliver on the full potential of AI. We’re excited that Ayar Labs is one of the first deployments that leverages the UCIe platform to its full extent.”

This sentiment was echoed by Kevin Soukup from GlobalFoundries, who noted, “As the industry transitions to a chiplet-based approach to system partitioning, the UCIe interface for chiplet-to-chiplet communication is rapidly becoming a de facto standard. We are excited to see Ayar Labs demonstrating the UCIe standard over an optical interface, a pivotal technology for scale-up networks.”

Technical Advantages and Future Applications

The convergence of UCIe and optical interconnects represents a paradigm shift in computing architecture. By combining silicon photonics in a chiplet form factor with the UCIe standard, the technology allows GPUs and other accelerators to “communicate across a wide range of distances, from millimeters to kilometers, while effectively functioning as a single, giant GPU.”

The technology also facilitates Co-Packaged Optics (CPO), with multinational manufacturing company Jabil already showcasing a model featuring Ayar Labs’ light sources capable of “up to a petabit per second of bi-directional bandwidth.” This approach promises greater compute density per rack, enhanced cooling efficiency, and support for hot-swap capability.

“Co-packaged optical (CPO) chiplets are set to transform the way we address data bottlenecks in large-scale AI computing,” said Lucas Tsai from Taiwan Semiconductor Manufacturing Company (TSMC). “The availability of UCIe optical chiplets will foster a strong ecosystem, ultimately driving both broader adoption and continued innovation across the industry.”

Transforming the Future of Computing

As AI workloads continue to grow in complexity and scale, the semiconductor industry is increasingly looking toward chiplet-based architectures as a more flexible and collaborative approach to chip design. Ayar Labs’ introduction of the first UCIe optical chiplet addresses the bandwidth and power consumption challenges that have become bottlenecks for high-performance computing and AI workloads.

The combination of the open UCIe standard with advanced optical interconnect technology promises to revolutionize system-level integration and drive the future of scalable, efficient computing infrastructure, particularly for the demanding requirements of next-generation AI systems.

The strong industry support for this development indicates the potential for a rapidly expanding ecosystem of UCIe-compatible technologies, which could accelerate innovation across the semiconductor industry while making advanced optical interconnect solutions more widely available and cost-effective.

[techdriveplay]

Why Quantum Computing Will Change the Tech Landscape

The technology industry has seen significant advancements over the past few decades, but nothing quite as transformative as quantum computing promises to be. Why Quantum Computing Will Change the Tech Landscape is not just a matter of speculation; it’s grounded in the science of how we compute and the immense potential of quantum mechanics to revolutionise various sectors. As traditional…

[dallasstarsdyke]

talking to myself is so fucked up when im hilarious. what if im getting a good grade in vocal stimming and no one hears

[1jet2unknown]

Did I just write 2000+ words of nothing but Chen Yi and Ai Di fucking?

Yes.

You bet your ass I did.

[d-the-designer]

[seancosy]

What frustrates me about AI is that it will be wasted on furthering corporate and nationalistic interests, rather than being used to unite humanity and improve our quality of life.

Imagine a world where AI is used to allocate resources, predicting surpluses and shortfalls, and considering transport times between communities, so that every region on earth has enough staple foods to mitigate starvation.

Imagine AI predicting the upcoming weather patterns, allowing farmers to know how much water and fertiliser are required for the coming season, or allowing people to escape superstorms/fires/floods well before the weather has even turned bad.

Imagine AI designing farms based on the most efficient design for a specific landscape - rural or urban - to maximise harvests and minimise waste.

Imagine AI predicting the likelihood of someone’s health deteriorating based on their vital signs and blood tests, and recommending effective treatments before the patient’s health declines too far.

Imagine AI designing tailored clothes, prosthetics, wheelchairs, and other functional/mobility aids, massively increasing accessibility for everyone.

Instead, we’re in a world of stolen art, surveillance, falsified academia, and deepfakes. We’re more likely to see fake footage of like... Xi Jinping saying a slur, than we are to end poverty within our lifetimes. And that’s a tragedy.

[stormyisqueer]

HELL YEAH!!!!!!!!!!

[certivo]

Making CE Product Compliance Easy: How Manufacturers Can Streamline EU Certification with Alex Volf

In this episode of Supply Chain Certified by Certivo, host Kunal Chopra talks with Alex Volf—founder of CE EASY and a leading expert in CE product compliance. With over 20 years at Bosch, Parker, and Hendrickson, Alex reveals how manufacturers can simplify CE marking, avoid compliance delays, and speed up time-to-market. Discover the role of AI in transforming certification processes and fixing supplier bottlenecks.

[aidrivenprototyping]

ProtoBoost.ai: Powering Rapid AI-Driven Prototypes & Product Launch Accelerator

ProtoBoost is an AI-driven prototyping platform that accelerates product development by transforming ideas into functional prototypes. Using advanced AI algorithms, it streamlines brainstorming, design, and visualization, ensuring rapid prototyping with AI. As an AI prototype generator, ProtoBoost enhances AI-powered product innovation, making concept-to-prototype transitions faster, smarter, and more efficient.