🤖 Jason Huggins

In today's episode of "Misinformation for the Masses," Linas invokes Moore's Law with the grace of a greased monkey juggling bowling balls. TL;DR: Moore's Law hasn't coincided with reality since the 1990s. This means that Moore's Law hasn't been relevant at any point in Linas' career considering Moore's Law was irrelevant before Linas learned to tie his own shoes. In 1965, Gordon Moore, co-founder of Intel, observed that the number of transistors on a microchip doubles approximately every two years, effectively doubling CPU performance. In the early 2000s, Moore's Law fell apart. We were already reaching the fundamental limits of how small a transistor can be and still perform reliably. Between 2000 and 2005, we saw the industry shift from the top-line speed of a single CPU core to adding multiple CPU cores to a single machine. In 2005, we saw the first consumer-grade, multi-core CPUs from Intel and AMD. In fact, I purchased my first multi-core CPU, an AMD Athlon 64 X2, in 2005 when I built a new gaming rig designed specifically for playing Half-Life 2. Between 2005 and today, the focus in computing has not been on the density levels of transistors in either CPUs or GPUs. The focus instead has been on parallel computing (more cores) and lower power consumption. One of the limiting factors in parallel computing is the nature of the underlying software and algorithms. There isn't much that can be done to make a browser or other desktop applications take advantage of multiple CPU cores. However, when it comes to rendering graphics and processing AI, the algorithms are natively parallel. This is why we have seen GPUs with hundreds of cores for over 20 years, while even today we don't see CPUs with more than 20 or 30 cores. For reference, AMD and Nvidia's top-line GPUs have almost 20,000 cores. Even Intel GPUs, which are optimized for low-power and low-cost, have hundreds of cores. Further problems with Linas' analysis here is that Blackwell is a cluster, not even a single GPU. The individual GPUs within the Blackwell cluster only have hundreds of TFLOPs. This is on par with AMD GPUs of the same enterprise class. What really sets Nvidia apart is not the raw performance. It is the CUDA platform, which makes it easy to run AI workloads on an Nvidia GPU. Nvidia released this in 2007 and has been maturing it ever since. CUDA is a boat that both Intel and AMD missed and are now struggling to close the gap on. Intel and AMD eventually realized this miss, but their efforts to develop/promote oneAPI and ROCm, respectively, have been lackluster. In summary, not only is Nvidia's lead not related to Moore's Law, it isn't even related to hardware performance. It is the culture they built on top of otherwise par hardware. That is why I invest in Nvidia. The community they built is what counts. Are you focused too much on your product, and not enough on the tools and community to support it? #Technology #Leadership

14

16 Comments

You need to be paying attention to Ohio tech. Here are some recent headlines... • Microsoft expands in Ohio with $420 million New Albany Data Center • Path Robotics raises $100 million Series D to drive AI-enabled robotic welding growth (h/t Andrew Lonsberry and Alex Lonsberry) • Over 100 Ohio CEOs unite to champion computer science education in new statewide initiative (Code.org, Julia Wynn) • Gifthealth's massive expansion to bring 751 new jobs to Columbus (h/t One Columbus) • Ohio invests $6.5 million in next-generation technology innovation (h/t Lydia Mihalik) • Ohio announces new funding round for tech startups through Third Frontier Program (Ohio Department of Development) • Transportation Research Center Inc. breaks ground on new Advanced Impact Laboratory (h/t Brett Roubinek) • Protecting Ohio: How local governments can better manage cyber risk and response (h/t Kirk Herath and Daniel Poliquin) • Augment Therapy partners with Niantic, Inc. to revolutionize physical rehabilitation with AR pets (h/t Lindsay Watson PT, MPT) • Ohio VC Fest 2024 sets new records, showcasing state’s growing startup opportunity (h/t Julie Jacono, JumpStart Inc.) • Toledo entrepreneurs create AI-powered, real-time support for mental health issues (h/t Sean Wheelock, Inner Peak AI (Techstars '24)) • Ohio Tech Day presented by Meta launches with immersive tech experiences for the next generation of innovators (h/t Tara Tenorio, Liz Morrison) • Ohio’s High School Tech Internship Program hits record numbers, expands for 2025 (h/t Geoff Smith, Keith Koehne) • Homeowner assist platform LiveEasy acquired for $80M by California firm • Demand for computer science graduates growing in Ohio • First ever Ohio AI Summit presented by Further is coming to Columbus on November 20th - and is now sold out (h/t Cal Al-Dhubaib) -- Link to each story in the comments. Follow OhioX and OhioTechNews.com for more tech news and events.

623

45 Comments

Anthropic's Computer Use feature for Claude is amazing . I just watched Alexander Finn's excellent demo video that shows how Claude can now interact with computer interfaces through prompts. Key highlights from the demo: -Claude can view and control computer interfaces - Enables automated task completion (base for AI agents systems) - Could transform how we handle tasks like debugging and customer support

7

1 Comment

'Smart' Contracts must be the dumbest name in #web3. Why? Well, the coding architecture that bears this name isn't particularly 'smart'. Smart Contracts are predominantly found in the 'EVM Compatible' Ecosystem. What is it? It is a (centralized) 'community of cryptographic architectures' that revolve around Ethereum. 'EVM' stands for 'Ethereum Virtual Machine'. Some of these architectures are blockchains, some aren't. From those that are not, some are transparent they are not, while the open information some provide about themselves are less clear. Many of the 'EVM Compatibles' have a link to Ethereum itself, but some don't. They use a common coding system, and this assists with interoperability. For example, it is very easy for a wallet provider, like MetaMask or Coinbase, to adjust to accommodate new 'EVM Compatibles'. A wallet designed to accommodate HNS Coins, Handshake Names/Domains, Bitcoin Ordinals, Bitcoin Stamps, Dogenals, and many other types of Web 3 related assets can't. Different types of addresses, identities and codes related to 'EVM Compatibles' have a common indicator in that they begin with the two characters '0x'. Most promo spiel on 'EVM Compatibles' will champion interoperability as if it being 'a good thing' is a given. It is not. Interoperability is a gateway, and if that gateway is open for joint operation between two or more architectures, it is also open to intrusion by bad actors. 9ja Cosmos uses this phenomenon to great effect in the 'Sinosignia' project because the single asian alike character (example: 𨰏 ) integral to the Handshake #blockchain and the Sino Amazon (tokenized artwork) cannot both inhabit the same wallet. We also have the '.det0x' names tokenized to #optimism while the 'Root Zone' is on Handshake. (Aside: if something claims to be a Web 3 Domain, check that it's 'Root Zone' is integrated through a blockchain core rather than secured through a Smart Contract. Some are not secured at all! If a provider is unclear about how it's Root Zone is secured, or doesn't mention it at all, that's not a good sign). 'EVM Compatibles' are not useless. but they are not decentralized. They are dangerous to use on their own when attempting to secure or 'mint' something collectible, and/or of tradeable value. Securing these kinds of assets with Smart Contracts is very risky. It can lose you those assets, and/or contribute to your identity to be stolen. $12.3 BILLION IN LOSSES — THAT’S THE AMOUNT RECORDED DUE TO SMART CONTRACT VULNERABILITIES GRACIOUS JOHN

23

9 Comments

Technology fails This one gave me a certain amount of schadenfreude...all the honking you'd expect on a normal day in traffic in New York without the flurry of extended middle fingers...LOL - https://lnkd.in/gfSpFeWw #ai #technology #mobility #waymo #thefundraisingangel MONEYWATCH Waymo's driverless cars honk at each other, waking neighbors By Megan Cerullo Edited By Anne Marie Lee Updated on: August 14, 2024 / 3:21 PM EDT / CBS News Waymo driverless cars are coming alive at night, honking at each other and waking residents around the San Francisco parking lot where they sit when not making trips. The driverless taxi company has rented a lot outside of an apartment building for the cars to idle in, but the vehicles appear to be honking at each other, as they are trained to do on the road when other cars get close to them. Randol White, who witnessed the honking activity, told CBS News that he heard "a pair of honks" on Wednesday around 4:30 a.m. Pacific time. He added that according to his neighbors, "peak honk" took place around 4:00 a.m. Waymo attributed the disruptive honking to a feature designed to avoid collisions. "We recently introduced a useful feature to help avoid low-speed collisions by honking if other cars get too close while reversing toward us," the company said in a statement to CBS News. "It has been working great in the city, but we didn't quite anticipate it would happen so often in our own parking lots." Waymo added that it has updated the software that causes the cars to honk "so our electric vehicles should keep the noise down for our neighbors moving forward." Video posted by White of more than two dozen cars all in the same lot, with some honking at each other, has circulated on social media. In his caption, White commented that he would have found the scene funny if it weren't taking place at 4 a.m. Software engineer Sophia Tung has set up a livestream of the lot on YouTube, telling The Verge that she likes watching the cars "come and go" but wishes they would stop honking at each other. Waymo's taxi service is currently available to anyone who download's the company's app in Phoenix, San Francisco and Los Angeles.

8

1 Comment

Hahahahahaha... oh my. The Verge: "Humane’s daily returns are outpacing sales/ The company is scrambling to stabilize as it hits $1 million in total returns against $9 million in sales." The Verge: https://lnkd.in/gRW28Hqr ----- Engadget: "People are returning Humane AI Pins faster than the company can sell them. Leaked internal sales documents suggest it has yet to rebound from a wave of bad reviews." (Lawrence Bonk) (Aug 7, 2024) Humane’s universally derided AI Pin is not exactly flying off of store shelves, according to internal sales documents published by The Verge. As a matter of fact, returns seem to be currently outpacing new purchases. So many have been returned, that there are only around 7,000 units out in the wild. The company had once hoped to sell 100,000 in the first year, so that ain’t happening. To date, around 10,000 have been sold and 3,000 have been returned, leaving the aforementioned 7,000 in the hands of consumers. This gadget costs $700, so I’d obviously have returned it after learning it doesn’t actually do anything useful. Wouldn’t you? To that end, over 1,000 pre-orders were canceled once reviews started coming in. We called it the “solution to none of technology's problems” and struggled to understand who it was even for. This was before the charging case accessory became a potential fire hazard, which added to the image of the product being the ultimate example of tech hubris. There’s a reason, after all, why noted tech vlogger Marques Brownlee called it “the worst product” he had ever reviewed. Humane has been, sort of, trying to turn things around, mostly by trying to find a larger company to scoop it up into its portfolio. HP has been considering a purchase, according to The New York Times. Also, the company’s allegedly been in active negotiations with investors, as reported by The Information. Humane has taken umbrage with The Verge and its reporting, with spokesperson Zoz Cuccias claiming “inaccuracies” regarding the financial data but didn’t go into specifics. Instead, she said “we have nothing else to provide as we do not comment on financial data, and will refer it to our legal counsel.” The Verge reports that the company has no refurbishment operation in place, which is another lost opportunity for revenue. The problem allegedly boils down to a limitation with carrier partner T-Mobile that doesn’t allow Humane to reassign one of these pins to a secondary user. The returned pins could be destined for the scrap heap, but Humane is reportedly holding onto the returns for now, in the hopes it solves the problem with T-Mobile. The company has been releasing software updates to address user feedback, so who knows. Maybe the ship can right itself. There’s still the question of spending $700 for a secondary device that underperforms when compared to a smartphone. The same question applies to a $200 device. We’re looking at you Rabbit." Engadget: https://lnkd.in/gfhAc6K5 #ai #fakeai #genai #aiwearable #aipin

51

8 Comments

I believe 3 categories of people and companies would benefit the most from automating image/video generation with Stability AI's Stable Diffusion and ComfyUI. I list each of them and their motivations in this video: https://lnkd.in/eaCh-FsR #comfyui #stablediffusion #sd3 #apw #ai #automation

2

Are we addicted to feeding the slot machines of bad content? The flood of bad content mastered exploitation of our short attention span and knows exactly how to fiddle with our dopamine levels. Video/news feeds aka "slot machines" promise a quick win, but rarely bring any substance. In fact, it's US who is feeding that machine. The urgency to make everything quick and digestible created a market craving of easy half-truths and self-adoration. Years of sorcery boil down to a clickbait title. We somehow forgot that growth is the messy process with not so easy to understand contexts - and we wish for shortcuts everywhere. We create the demand in this market. We reward those convenient lies because they are easier to consume than the messy truths. Stepping away (sick & full calendar) made it clear how much noise we're exposed to - and how easy it is to feed into it. I challenge you: be intentional about what you consume. Pick authors that not necessarily have huge following, but are committed to creating meaningful content. They might struggle with consistency and they won't always deliver a perfect copy. But they STILL CARE. And that makes them worth following. Because when people demand the impossible, only liars will satisfy them. ---- 🔔 If you are tired of noise and crave some balance and real-world perspectives - follow me. ♻ Share. That is they only way I can break through the algorithm's gate... 💡 Check the comments to see who inspired this post... 🖼 It's me. One of the very few moments I can remember from the final ascent to Kilimanjaro. Lack of oxygen and basically "no acclimatization" reduce your abilities to the absolute minimum. At that point I did not know it - but minor frostbites on my toes were already "ready". Yep, another mistake in planning. This climb wasn't just a test of my personal limits and organization - it was about trusting the process and relying on the team supporting us on the way to the roof of Africa. It was messy. It was hard. Brutal and humbling. You can't capture it in 3 sentences. And you can't make it up.

10

12 Comments

Just came across a fascinating post about Python’s watchdog library enabling easy monitoring of your file system and unlocking the potential of automation. With this in place, you can now watch over your file system in real time, making necessary adjustments on the fly. This cross-platform API accelerator allows us to attain cutting edge technology right from our own devices or workspaces. • The beauty of watchdog rests in its ability to allow commands based on activity in the file system being looked at. • You can set triggers around multiple events, creating responsive and adaptive systems. This not only exhibits the core versatility and adaptability of Python’s library but promises tremendous implications for business automation and operational efficiency. On a predictive note, APIs such as the watchdog will perhaps drive the next wave of innovation in operational optimization and workflow automation. It will likely unearth new methods to run intelligent systems by leveraging real-time sentiment changes and their effective incorporation. In an era graced by digital disruption and technological advancement, such compelling offerings amplify the importance of adaptive learning frameworks and quick decision-making processes, thrusting them into the limelight. This initiative from Python resumes the push for operational excellence via practical and purposeful technology. Overall, the prowess of watchdog echoes the accompanied transformational journey of today’s developers and technology leaders alike. So, gear up to monitor, adjust and hit your targets using Python's watchdog and let automation become your organization's strategic pillar! #Python #WatchdogLibrary #Automation #FileSystemMonitoring #APIs #TechInnovation

1

Web3 security starts with hands-on practice. Want to level up your skills? Check out my Ethernaut + Foundry playlist—free resources to get you started! No excuses!

16

New Capabilities of Gemini 1.5 Flash: Google’s response to GPT-4o Gemini 1.5 Flash has emerged as a worthy contender for OpenAI’s much hyped GPT-4o. Here's what the community is saying: - Unmatched Speed: Gemini 1.5 Flash is redefining efficiency with its lightning-fast processing, making it a go-to for tasks that demand quick turnaround times. - Cost-Effective: Balancing performance with affordability, Gemini 1.5 Flash is proving to be a smart investment for those seeking powerful capabilities without breaking the bank. - Impressive Reasoning: While it may not boast the raw power of its Pro counterpart, Gemini 1.5 Flash holds its own with robust reasoning abilities, perfect for rapid content generation. - Extended Context Window: The model's expansive context window allows for comprehensive interactions, setting a new standard for consumer AI chatbots. - Multimodal Mastery: With its adept handling of text, images, and videos, Gemini 1.5 Flash is a testament to the potential of multimodal AI. - User Interface: Despite some feedback on its interface, the model's strengths in speed and efficiency shine through, marking it as a valuable tool for developers and users alike. Gemini 1.5 Flash will soon be joining Swarms framework compatibility lists. Stay tuned for more! #AI #Gemini15Flash #ArtificialIntelligence #Innovation #Technology #Google #OpenAI #Swarms

3

Drone deliveries (AI- enabled drones) are taking off. Users include hospitals like Cleveland Clinic and restaurants like Jets Pizza and Panera Bread. “(Drones) are light, fast, autonomous and zero-emission.” “Delivering ... is 10 times as fast, it’s less expensive … and relative to the traditional delivery apps that most restaurants work with." "We triple the service radius, which means you actually [get] 10 times the number of customers who are reachable via instant delivery.” And yes, Zipline uses Edge AI and robotics platform to power its drone fleet. The AI system-on-module enables the drones to navigate autonomously. AI also plays a key role in Zipline's delivery system, helping with route planning, demand prediction, and maintenance. AI-enabled drones use machine learning and computer vision to analyze data, recognize patterns, and make real-time decisions. This allows the drones to navigate autonomously, detect objects, and make decisions. Go Zipline. https://lnkd.in/dVp8WDzX Mike Edwards Peter Cobb Brian Sathianathan Magnus Tagtstrom Dave Jenkins

16

1 Comment

Every ChatGPT query uses a gulp of water. An ordinary session can consume the equivalent of a 16-ounce bottle of water (for cooling servers at data centers). So, how are tech companies going to meet AI's growing thirst? I just published an article in Scientific American with Bettina Warburg making the case that tech firms should massively scale up rainwater harvesting at data centers. If harvesting rainwater sounds simple, that’s because it is.  People have been gathering water from the sky since ancient times. Today, there’s growing interest in the technology of collecting rain from rooftops and sending it down gutters into tanks or retention ponds for later use.  A range of recent studies show that harvesting even a small portion of the rainwater that falls in a given area can eliminate water shortages while simultaneously recharging groundwater and reducing pollution from stormwater runoff. It’s not just that data centers have a pressing need for massive amounts of water—it’s also that their large flat roofs can harvest a lot of what falls. A 50,000 square foot roof can collect about 31,000 gallons of water from a single inch of rain.  Much like installing solar, installing rainwater harvesting capacity is a one-time investment that cuts long-term utility costs—while meeting other goals like resilience and conservation. Thanks to Israel Mirsky and Erin English, PE LEED AP for help with this piece. Here's a link: https://lnkd.in/giByqEWQ #water #climate #AI #datacenters #climateadaptation #waterscarcity #watercrisis #rainwaterharvesting #watertech #drought #waterconservation #climatetech #greentech Rainwater Harvesting Ltd IRHA - International Rainwater Harvesting Alliance rainwater Rainwater Management Solutions Stormsaver Ltd Rainwater Collection Tanks Willdan Performance Engineering Biohabitats The UAE Research Program for Rain Enhancement Science (UAEREP) Steven Hudgins Gravity Water

74

6 Comments