Bill Kallman

MSR has showcased BitNet, a novel 1-bit Transformer architecture, designed to address the challenges associated with scaling LLMs. A key innovation in BitNet is the introduction of BitLinear, which replaces traditional linear layers and facilitates the training of 1-bit weights from the ground up. This architecture not only achieves competitive performance in language modeling but also significantly reduces memory usage and energy consumption when compared to advanced 8-bit quantization methods and FP16 Transformer baselines.   The architecture employs quantization-aware training, optimizing the model for low-precision representation from the outset, which enhances both accuracy and stability during the training process. BitNet features stacked blocks of self-attention and feed-forward networks with binarized weights, while retaining higher precision for other components, thereby minimizing computational costs. This design allows for efficient matrix multiplication with 1-bit weights, resulting in notable energy savings, especially as model sizes increase.   The training methodology incorporates techniques such as the straight-through estimator for gradient approximation and mixed precision training. These approaches enable BitNet to converge effectively with larger learning rates compared to traditional FP16 Transformers, leading to improved training efficiency and stability, which is particularly beneficial in resource-constrained environments.   Comparative evaluations indicate that BitNet consistently outperforms existing quantization methods across a range of benchmarks, including zero-shot and few-shot tasks, while also maintaining lower inference costs. These findings highlight the advantages of quantization-aware training over post-training quantization, positioning BitNet as a prominent approach for developing efficient LLMs.    Arxiv: https://lnkd.in/e-D959-U

3

New Tesla Patent Reveals Clues to an Exciting Future of Humanoid Robots and AI Integration In a interesting discussion, Herbert Ong and Scott Walter dive deep into Tesla's recently published patent application WO2024072984A1 (link in comments). Entitled "Actuator and actuator design methodology," the patent primarily focuses on controlling robot movement through strategic actuator placement. However, the document also unveils important clues about the potential for end-to-end neural net training of humanoid robots, reminiscent of Tesla's approach to training its vehicles with FSD. Ong and Walter explore the concept of personalized humanoid skills. Imagine purchasing a bot with built-in capabilities, then having the option to augment and customize those skills through a "Skills Store" akin to Apple's App Store. This opens up a world of possibilities for tailoring robots to specific tasks and preferences. The discussion also touches on the potential for various robot form factors, each designed to excel at particular skills. Picture a tennis-playing robot that comes pre-trained with basic volleying instructions, but requires a "Skills Upgrade" to perform at a master level. This raises questions about the need to manufacture and train a diverse range of robot body types and sizes, mirroring how humans possess ideal physiological characteristics for different activities. Later, Ong and Walter discuss a scenario in which Tesla builds the hardware – including automobiles, humanoids, and other bot form factors – while xAI develops the brains. This arrangement opens up opportunities for licensing agreements between the two companies, that may help the companies navigate future antitrust concerns as these technologies are deployed at scale. The convergence of advanced robotics and artificial intelligence holds the potential to catalyze a profound societal transformation. As Tesla and xAI continue to push the boundaries of what's possible, I can envision a future where intelligent, adaptable robots become integral to various aspects of our lives – from manufacturing and transportation to healthcare, recreation, and beyond. The seamless integration of cutting-edge hardware and software will revolutionize the way we live and work, ushering in an era of unprecedented efficiency, productivity, and innovation. While there are still challenges to overcome and ethical considerations to address, the prospect of these combined technologies driving positive change on a global scale is awe-inspiring. #FrontierFrontier #Humanoids #AI #Robots

7

1 Comment

Really interesting to see this: how Europe is moving up. It is interest to contrast this qualitative survey with the #siliconcatalyst analysys that UK is sixth in Europe in funding value fir silicon investment. And I will put in a plug for our #UKTIN Forward Capabilities paper on recommendations for UK Semiconductor sector: https://lnkd.in/eNNAsCAz

8

One of the hidden reasons why AR/VR has not taken off is that people get nausea as our eyes are not accustomed. Now there is a game-changing technology available and we are proud to back it.

44

3 Comments

September 02 2024 – 5 Notable Articles To Read Article: Chip challengers try to break Nvidia’s grip on AI market Takeaway: Cerebras Systems Inc., d-Matrix, and Groq are part of a group of smaller companies seeking to capture a portion of the multibillion-dollar AI chip market, which Nvidia has largely dominated with its graphics processing units (GPUs) during the initial wave of investment. Source: Financial Times Link: https://lnkd.in/d8yh-VVU Article: ‘T-Bill and Chill’ Is a Hard Habit for Investors to Break Takeaway: Money-market funds surge to a record $6.24T, and while upcoming Fed rate cuts suggest a shift may be warranted, the timing remains uncertain. Source: Bloomberg Link: https://lnkd.in/dp6wvceH Article: Why Costco Wholesale Could Be the Next to Split Its Stock Takeaway: The price increases are likely a key factor driving the stock splits, which make shares more accessible to retail investors and, some believe, psychologically more appealing to potential shareholders. This context highlights Costco, which has surged nearly 40% since the start of 2024, with its last split occurring in 2000, just a year after Walmart’s previous split. Source: Barron's Link: https://lnkd.in/dyPpVbAq Article: Meta Shuttering Augmented Reality Studio Amid Pivot to AI Takeaway: Meta Spark was originally intended to compete with Snapchat and the company will shift resources to products such as smart glasses. Source: Bloomberg News Link: https://lnkd.in/daEY47n7 Article: AI Regulation Is Coming. Fortune 500 Companies Are Bracing for Impact. Takeaway: Approximately 27% of recent filings with the Securities and Exchange Commission cited AI regulation as a risk, marking one of the clearest indications yet of how AI rules could impact businesses. Source: The Wall Street Journal Link: https://lnkd.in/dS977wWK #news #research #business #ai

6

Arcee AI Introduces Arcee Agent: A Cutting-Edge 7B Parameter Language Model Specifically Designed for Function Calling and Tool Use https://lnkd.in/ef-2vS_G #llm #ai

1

The Cerebras IPO looking to break up the NVIDIA monopoly might be successful in the short-term for an IPO pop, but long term will likely fail. NVIDIA’s dominance doesn’t come just from its chip but from the CUDA software that is the de facto standard and can take advantage of chip architecture improvements. No wonder why 87% of Cerebras’ revenue is from the Middle East. The people buying it aren’t the engineers using it. https://lnkd.in/ePRqdFpt

4

3 Comments

𝗦𝗲𝗺𝗶𝗰𝗼𝗻𝗱𝘂𝗰𝘁𝗼𝗿 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗩𝗲𝘁𝗲𝗿𝗮𝗻𝘀 𝗟𝗮𝘂𝗻𝗰𝗵 𝗔𝗸𝗲𝗮𝗻𝗮 𝗮 𝗥𝗜𝗦𝗖-𝗩 𝗜𝗣 𝗰𝗼𝗺𝗽𝗮𝗻𝘆 𝗮𝗻𝗱 𝗥𝗮𝗶𝘀𝗲 $𝟭𝟬𝟬𝗠 The golden age of semiconductors continues and has only been amplified in the GenAI era. This wave has guided some of my investments including Alif Semiconductor, Auradine, Frore Systems, Fungible Inc., Acquired by Microsoft, NUVIA Inc (acquired by Qualcomm), Recogni, SambaNova Systems, and SiMa.ai. I’m delighted to announce the formal launch of our portfolio company Akeana, a RISC-V International IP company that was founded three years back and has raised $100M since then. We are excited to partner with an exceptional team of industry veterans, including Rabin Sugumar, Nitin Rajmohan, Abbas Rashid, and Bruno Patnam, who bring a wealth of experience from companies including Arm, Broadcom, Cavium Inc, Marvell Technology, Intel Corporation and Sun Microsystems/Oracle. Akeana builds upon the groundbreaking work of MIPS, a fabless semiconductor company we invested in back in 1984, known for pioneering the RISC architecture. MIPS was founded by Dr. John Hennessey, ex-President of Stanford University and current Chair of the Board of Google who along with David Patterson, Professor Emeritus at the University of California, Berkeley pioneered work in the RISC space . Akeana announced multiple products including: 𝗔𝗸𝗲𝗮𝗻𝗮 𝟭𝟬𝟬 𝗦𝗲𝗿𝗶𝗲𝘀: A line of highly configurable processors with 32-bit RISC-V cores that supports applications from embedded microcontrollers to edge gateways, to personal computing devices. 𝗔𝗸𝗲𝗮𝗻𝗮 𝟭𝟬𝟬𝟬 𝗦𝗲𝗿𝗶𝗲𝘀: A processor line that includes 64-bit RISC-V cores that support multi-threading, vector extension, hypervisor extension for applications in wearables, TV and mobile devices. 𝗔𝗸𝗲𝗮𝗻𝗮 𝟱𝟬𝟬𝟬 𝗦𝗲𝗿𝗶𝗲𝘀: A processor line of 64-bit RISC-V processors representing industry performance leadership for demanding applications in laptops, data centers, and cloud infrastructure. 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗼𝗿 𝗦𝘆𝘀𝘁𝗲𝗺 𝗜𝗣: A collection of IP blocks needed for creation of SoCs, including Coherent Cluster Cache, I/O MMU, Interrupt Controller, Scalable Mesh and Coherence Hub IPs.   𝗔𝗜 𝗠𝗮𝘁𝗿𝗶𝘅 𝗰𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻 𝗲𝗻𝗴𝗶𝗻𝗲: Designed to offload Matrix Multiply operations for AI workloads for training and inference in data centers, cloud, and edge devices. Looking forward to partnering with many more semiconductor entrepreneurs in their inception to iconic journeys, especially in the #GenAI space. #RISCV #GenAI #Semiconductors #Founders 

149

7 Comments

𝙒𝙝𝙖𝙩 𝙘𝙖𝙪𝙜𝙝𝙩 𝙢𝙮 𝙖𝙩𝙩𝙚𝙣𝙩𝙞𝙤𝙣? AI headlines have spotlighted nuclear-powered data centers, but yesterday, Pacific Gas and Electric Company signed with Atomic Canyon to deploy AI software that will streamline document retrieval, cut costs, and boost operational efficiency, supporting its license extension efforts at the Diablo Canyon Power Plant . 𝙒𝙝𝙮 𝙞𝙩 𝙢𝙖𝙩𝙩𝙚𝙧𝙨? AI-driven software that extends the operating life of physical assets like Diablo Canyon—a large-scale, low-carbon, and dispatchable generation source—plays a key role in meeting the rising energy demands, much of which are fueled by AI itself. This is a good example of software’s potential to address the grid’s physical limitations. Other promising applications include software that optimizes transmission infrastructure, manages virtual power plants, and enhances load flexibility. 𝘿𝙚𝙩𝙖𝙞𝙡𝙨: • Diablo Canyon generates approximately 2,300 megawatts (MW) of electricity, and represents about 9% of California’s total electricity generation. • The facility holds approximately 9,000 procedures and 9 million documents, many digitized from paper or microfiche. To extend its federal license for up to 20 more years, PG&E must develop aging management plans, relying heavily on decades-old records. • PG&E's noted that AI could eventually assist with complex tasks like maintenance scheduling, which requires understanding the plant's interconnected systems. This labor-intensive process could see significant efficiency gains through AI. For updates on power markets, renewables, carbon capture, and the broader energy transition, follow me. Enverus 𝘗𝘰𝘸𝘦𝘳𝘪𝘯𝘨 𝘵𝘩𝘦 𝘎𝘭𝘰𝘣𝘢𝘭 𝘘𝘶𝘢𝘭𝘪𝘵𝘺 𝘰𝘧 𝘓𝘪𝘧𝘦 #nuclear #datacenters #AI #grid #energy

6

Silicon Valley startups are finding their way into government contracting opportunities as cheaper military drones become a bigger problem worldwide. Anti-drone technologies take numerous shapes and sizes, but the true potential of what is possible has yet to be explored. In the coming years, I believe we can expect to see even more intelligent uses of drone tracking technology to help face the threat that unmanned drones present in combat theatres worldwide. #Drones #Innovation #Startups

2

🔬 𝐍𝐞𝐰 "𝐆𝐨𝐥𝐝 𝐒𝐭𝐚𝐧𝐝𝐚𝐫𝐝" 𝐢𝐧 𝐌𝐢𝐜𝐫𝐨𝐬𝐜𝐨𝐩𝐲 𝐃𝐚𝐭𝐚 𝐀𝐧𝐚𝐥𝐲𝐬𝐢𝐬 𝐰𝐢𝐭𝐡 𝐎𝐩𝐞𝐧𝐏𝐡𝐞𝐧𝐨𝐦 𝐛𝐲 𝐑𝐞𝐜𝐮𝐫𝐬𝐢𝐨𝐧 🔬 Utah-based Recursion has launched 𝐎𝐩𝐞𝐧𝐏𝐡𝐞𝐧𝐨𝐦, a non-commercial, AI foundation model that aims to revolutionize microscopy data analysis, claiming to outperform traditional tools like 𝐂𝐞𝐥𝐥𝐏𝐫𝐨𝐟𝐢𝐥𝐞𝐫. OpenPhenom-S/16 is adaptable, removing the need for constant fine-tuning—an essential feature as large-scale datasets like RxRx3 and JUMP-CP continue to expand. Foundation models are rapidly advancing #biotech! Curious about companies pioneering AI in pharma?  Stay tuned for more updates. BioPharmaTrend.com #AIDrugDiscovery #Microscopy #BiotechInnovation #Recursion #FoundationModels #DrugDiscovery

4

America may run out of power with AI's growing energy appetite 🔌🏭 The bottleneck for AI progress will shift from NVIDIA/TSMC's chip production to energy companies' capacity to power AI data centers. For example: ⚡️ Amazon's 1GW data center next to a nuclear plant https://lnkd.in/g4zqTwhG ⚡️ Microsoft & OpenAI planning $100B, multi-gigawatt "Stargate" AI supercluster https://lnkd.in/gWvnhZms To feed the AI beast, we'll need 💰💰💰 of investments in new power plants, massive data centers, and chip fabs that can churn millions of GPUs/year. Leopold Aschenbrenner's back-of-the-envelope analysis highlights the staggering power needs (https://lnkd.in/gfBc5xhb) 💡 Rumored $100B cluster = 10GW = small US State's power draw The intersection of AI and energy markets is 🔥🔥🔥 Think beyond chips and models. What's the NVIDIA or TSMC of the energy space? Nuclear? Solar? Exotic energy source? #ai #ml #llm #largelanguagemodels #gpu #cloud #datacenter #aicompute #energy #datacenterenergy #nuclear #greenenergy #startups #venturecapital

35

24 Comments

NVIDIA is putting the "silicon" back in "Silicon Valley." It has just become the most valuable company in the world. The stock is up almost 10x since the end of 2022. But if there's one thing you can count on in this world, it's that what goes up fast can come down just as fast. Don't get me wrong, Jensen deserves his Taylor Swift moment. The NVIDIA story is basically business legend. It's a story of perseverance, grit, ingenuity, and "anything's possible if you put your mind to it." And if you get really lucky. And there's good fundamentals behind the stock movement. Its latest quarterly earnings shows a net income of $15B, which is an annual run-rate of $60B, up 7x YoY. Given its growth rate, being valued at 50x annualized net income is almost a modest price to pay. Compared with meme stocks like GameStop, anyway. But the biggest difference between NVIDIA and Apple / Microsoft (which have traded the title of most valuable company between themselves over the last few years) is that Apple & Microsoft have built an incredibly sticky product that sits on top of the value stack. Their ecosystems are massively valuable, making switching away from them in the short term almost impossible. NVIDIA, on the other hand, has a big customer concentration problem. Roughly half of its revenues in recent quarters have come from Amazon Web Services and Microsoft, which have been installing GPUs in their data centers at a feverish pace to meet customer demand for AI workflows. Worse, consumers and the broader economy don't interface with NVIDIA at all, making developers' task of switching away from NVIDIA GPUs that much easier if better / cheaper / faster alternatives were to come along. Remember "Intel Inside"? They used to be the only way to go. But Intel's revenues have dropped 32% from 2021 to 2023 and that trend looks increasingly difficult to reverse. Finally, AI continues to be a massive investment that, so far, has generated little return. If companies decide to take a breather on their AI investment, that can mean a quick drop for NVIDIA's revenue. Of course, the opposite could be true. NVIDIA could continue to dominate the AI compute market. AI investment could continue its meteoric rise. NVIDIA's push into software and cloud services could solidify its position and increase the cost to switch. But whether the "NVIDIA AI Enterprise" (that's a mouthful) is gonna be successful is too early to tell, and certainly will face tons of competition in its fight to the top. Anyway, none of this is investment advice and is for informational and entertainment purposes only. 🙂 #NVIDIA

49

4 Comments

TSMC third-quarter profit seen jumping 40% on strong AI chip demand via Reuters The world's largest contract chipmaker is poised to report a staggering 40% increase in third-quarter profits, driven by surging demand in the AI sector. With major clients like Apple, NVIDIA, AMD, and Qualcomm launching new products that rely heavily on TSMC's advanced chip technologies, the company's net profit is expected to hit T$298.2 billion ($9.27 billion) for the quarter ended September 30. As TSMC continues to expand its production capabilities—investing billions in new factories, including $65 billion in Arizona—it's clear that the AI boom is significantly impacting its growth. With shares skyrocketing 77% this year, TSMC is leading the charge in the semiconductor industry. Link to article: https://lnkd.in/dAW3UqrY #TSMC #Semiconductors #AI #Earnings #TechIndustry

6

"Chinese artificial intelligence developers are using foreign computing resources to get access to cutting-edge American chips like Nvidia's A100 in a cunning move meant to evade strict US export regulations, according to The Wall Street Journal. These developers are getting around the restrictions placed on direct imports of these highly sought-after chips by collaborating with brokers and using identity-mapping strategies taken from the cryptocurrency sector. This workaround is a reaction to the US government's attempts to keep Chinese enterprises from obtaining the technology required for the development of cutting-edge artificial intelligence. Although it is still possible to physically smuggle Nvidia chips through grey-market channels, the procedure is frequently difficult and cannot satisfy the enormous demand of major Chinese AI companies. This new approach provides Chinese developers looking to gain access to more covert and effective" "According to the WSJ report, one entrepreneur helping Chinese companies overcome the hurdles is Derek Aw, a former bitcoin miner. He persuaded investors in Dubai and the US to fund the purchase of AI servers housing Nvidia's powerful H100 chips. In June, Aw's company loaded more than 300 servers with the chips into a data centre in Brisbane, Australia. Three weeks later, the servers began processing AI algorithms for a company in Beijing. "There is demand. There is a profit. Naturally, someone will provide the supply," Aw told the news portal. Renting faraway computing power is nothing new, and many global companies shuffle data around the world using US companies' services such as Google Cloud, Microsoft Azure, and Amazon Web Services. However, those companies, like banks, have "Know Your Customer" policies that may make it difficult for some Chinese customers to obtain the most advanced computing power. Listen to the latest songs, only on JioSaavn.com Also, lawyers familiar with US sanctions believe that there is no legal breach by buyers, dealers, or middlemen who move computing power."

4