AI comes to PCs at Computex 2024 - Weekly News Roundup - Issue #470
AI comes to PCs at Computex 2024 - Weekly News Roundup - Issue #470
Plus: US launches antitrust investigations into Microsoft, OpenAI, and Nvidia; Nvidia passes $3T valuation; Right to Warn about Advanced Artificial Intelligence; scaling CRISPR cures; and more!
Hello and welcome to Weekly News Roundup Issue #470. This week, computer hardware companies, including the likes of Nvidia, AMD, Intel, and Qualcomm, showcased their latest products at this year’s Computex conference. As with everything in the tech world recently, the main theme of Computex 2024 was AI.
In other news, the US has launched antitrust investigations into Microsoft, OpenAI, and Nvidia. Meanwhile, Nvidia has passed $3 trillion in valuation, becoming the second most-valued company in the world.
In robotics, 1X demonstrated what their humanoid robots can do, while a Chinese car manufacturer introduced humanoid robots into its factory.
I hope you enjoy this week’s issue!
Let’s start with AMD, which announced the launch of the next generation of Ryzen CPUs, the Ryzen 9000 series, based on the new Zen 5 architecture. AMD claims the CPUs deliver 16% better IPC on average compared to the previous Zen 4 generation. The new Ryzen 9000 CPUs are scheduled to be available in July and their price has not been revealed yet.
Alongside Ryzen 9000 CPUs, AMD also launched the new Ryzen AI 300 series of processors, designed to be used in laptops, which will come with a built-in Neural Processing Unit (NPU) to accelerate AI operations for the emerging class of AI PCs. AMD claims it is the most powerful NPU for PCs, delivering 50 TOPS (tera operations per second).
The first laptops with Ryzen AI 300 processors are expected to be available in laptops starting in July 2024.
In addition to the new processors, AMD also provided more details about the upcoming AMD Instinct MI325X accelerator to challenge Nvidia’s B200 Blackwell GPUs. The MI325X is expected to be released in Q4 2024. For more information about the MI325X, I recommend this in-depth analysis on AnandTech, which details what AMD is bringing to the table.
Intel also brought a new laptop CPU, named Lunar Lake, to Computex. Similar to what Apple has done with Apple Silicon chips, Lunar Lake will combine CPU, GPU, NPU, and memory into one chip to improve power efficiency. The new processor promises up to 14% faster CPU performance at the same clock speed as the previous Meteor Lake, 50% more graphics performance, and up to 60% better battery life than last year’s model. Lunar Lake will be Intel’s first CPU to feature an NPU offering up to 48 TOPS. Lunar Lake is expected to be available in late 2024. If you want to learn more about Intel’s new laptop processors, I recommend checking out The Verge’s article.
During his keynote, Pat Gelsinger, Intel CEO, also presented Intel's Gaudi AI accelerators. The company is positioning Gaudi 2 and Gaudi 3 as better AI accelerators for inference and more cost-efficient than Nvidia’s H100 GPUs, with Gaudi 3 delivering up to 40% faster training and 2x faster inferencing on popular models. Gelsinger also emphasized Intel’s commitment to open-source standards, taking a jab at Nvidia’s proprietary solutions like NVLink.
However, both AMD and Intel should be concerned with Qualcomm and the threat ARM-based CPUs can pose in the PC space. The launch of Apple Silicon in 2020 was a seismic event in computing, showing that light, energy-efficient and powerful ARM-based computers are possible and bringing a new golden age for Macs. Now, a similar change might be coming to PC, thanks to Qualcomm. Qualcomm’s chips are already powering the recently announced Microsoft’s new line of Copilot+ PCs and at Computex 2024, six more laptop manufacturers, Asus, HP, Dell, Acer, Samsung and Lenovo, unveiled their new AI PC laptops powered by Qualcomm’s Snapdragon chips. “It's the most disruptive thing in computing since Windows 95,” said Qualcomm CEO Cristiano Amon at a Q&A session a day after the keynote. Snapdragon X Elite chips integrate a CPU and GPU in one chip as well as a Hexagon NPU, which offers 45 TOPS for running AI applications.
Qualcomm’s entrance into the PC market could dramatically change the landscape, which is currently dominated by Intel and AMD. Rene Haas, the CEO of ARM, predicts that by 2029, ARM chips like the ones offered by Qualcomm will capture 50% of the Windows PC market. Qualcomm’s CEO, Cristiano Amon, shares a similar opinion. Qualcomm is starting with laptops, but it might soon branch out into mini-PCs and desktop PCs.
Meanwhile, Nvidia did not show any new GPUs at this year’s Computex. Some had hoped Nvidia would at least paper-release or tease the new RTX 5000 series of GPUs, but that did not happen. Instead, Nvidia focused on software and showed what can be done with their GPUs.
Nvidia is deeply invested in creating digital twins, or digital representations of physical objects and phenomena that can then be simulated in a virtual world (or Omniverse, as Nvidia calls it). At Computex, Nvidia presented Earth-2, a high-resolution climate and weather simulation. According to Jensen Huang, this digital twin of Earth can accurately simulate weather on every square kilometre of the planet.
For gamers, Nvidia has Project G-Assist. G-Assist is a chatbot trained to help gamers get unstuck. Nvidia hopes that instead of searching forums, Reddit, or Discord servers, gamers will ask G-Assist to guide them through games. I’m curious if G-Assist is a response to Microsoft’s demo of Copilot guiding how to play Minecraft. Additionally, G-Assist can also optimize PC settings to improve performance.
Game designers and developers could be interested in Nvidia ACE. Nvidia ACE is a suite of tools designed to generate digital avatars and characters in video games. It combines Nvidia’s expertise in computer-generated graphics with the latest advancements in generative AI. ACE can be deployed in the cloud or run locally on Nvidia’s RTX AI PCs.
With Nvidia ACE, game developers can create interactive NPCs that react to players' actions in real time. ACE can understand what the player has done and said, and then generate a unique interaction, including an AI-generated response, voice synthesis, and animation.
In addition to game development, Nvidia sees interactive AI-generated customer service agents as another application of ACE.
Although Nvidia did not show any new GPUs at Computex, Jensen Huang did not miss an opportunity to showcase their latest Blackwell systems. The new addition to the Blackwell family of products is Blackwell MGX, modular and liquid-cooled AI systems that can be tailored to different use cases. Jensen Huang also announced that Nvidia will release a new generation of "architecturally compatible" AI platforms every year. This year, Nvidia released the Blackwell line of GPUs. Next year, a refreshed Blackwell Ultra platform will be released, followed by the new Rubin architecture in 2026 and Rubin Ultra in 2027.
Nvidia’s keynote concluded with Jensen Huang inviting robots to the stage, saying that the next wave of AI is robotics.
This year’s Computex brought AI into PCs. Microsoft began the “AI PC” trend with the release of the latest Surface laptops and the introduction of a new class of Copilot+ PCs. Other PC manufacturers and chip companies followed suit, offering their own AI PCs and processors with built-in NPUs. The first AI PCs are expected to start shipping in July, and we will soon see how they work in practice and whether they actually make anything better.
Now let's see how Apple will respond at WWDC 2024 on Monday.
If you enjoy this post, please click the ❤️ button or share it.
Do you like my work? Consider becoming a paying subscriber to support it
For those who prefer to make a one-off donation, you can 'buy me a coffee' via Ko-fi. Every coffee bought is a generous support towards the work put into this newsletter.
Your support, in any form, is deeply appreciated and goes a long way in keeping this newsletter alive and thriving.
🦾 More than a human
Neuralink rival shatters world record, implants 4,096 electrodes in brain
Precision Neuroscience, a company building invasive brain-computer interfaces (BCIs), has implanted a record-breaking 4,096 electrodes in a human brain, doubling the previous world record of 2,048 set last year. The company used the Layer 7 Cortical Interface, a thin film microelectrode array engineered to conform to the brain’s cortex without damaging tissue. For more information about Precision Neuroscience and other BCI startups, check out my article "Six Startups Connecting Brains with Machines," which is part of my Brain-Computer Interfaces series of posts.
Gene therapy trial restores hearing in both ears for deaf children
Five children born deaf have gained hearing in both ears through a groundbreaking gene therapy trial conducted by doctors at Fudan University in Shanghai. The children, aged one to eleven, had genetic mutations affecting their ability to produce otoferlin, a protein essential for passing auditory signals from the ear to the brain. By delivering functional copies of the affected gene, Otof, to the inner ear, the children were able to hear, locate sound sources, and recognize speech in noisy environments within weeks. The trial marks a significant advance in gene therapy for genetic and potentially non-genetic causes of deafness.
▶️ Cryonics, cryocrastination, and the future: changing minds, with Max More (48:52)
In this podcast, London Futurists speaks with Max More—philosopher, futurist, transhumanist, and former CEO of Alcor. The conversation explores the early days of the transhumanism movement, how to think about the future, and how to regulate innovative technologies. The main topic is cryopreservation, discussing the scientific research supporting the idea and the barriers to its advancement, including the need for more research and how to overcome psychological arguments against cryopreservation.
🧠 Artificial Intelligence
US regulators to open antitrust inquiries of Microsoft, OpenAI and Nvidia, NYT reports
The US Justice Department and Federal Trade Commission (FTC) have agreed to open antitrust investigations into Microsoft, OpenAI, and Nvidia's dominance in the AI industry. The Justice Department will focus on Nvidia, while the FTC will investigate Microsoft and OpenAI. The FTC is also examining Microsoft's $650 million deal with AI startup Inflection AI. This agreement highlights increasing regulatory concerns over AI industry practices, with the FTC already investigating OpenAI for consumer protection violations and seeking information from major AI and cloud service companies about their recent investments and partnerships.
Nvidia hits $3tn and surpasses Apple as world’s second most valuable company
Nvidia just can’t stop growing. In 2024, the company’s stock surged 147% and hit an all-time high on Wednesday, crossing a $3 trillion valuation and becoming the world’s second most valuable company. The only company ahead of Nvidia in terms of valuation is Microsoft, which is valued at $3.14 trillion.
A Right to Warn about Advanced Artificial Intelligence
A group of current and former employees of OpenAI and Google DeepMind have launched the "A Right to Warn about Advanced Artificial Intelligence" campaign. They call on leading AI companies to support a culture of open criticism, stop using any agreements that prohibit “disparagement” or criticism of the company, and not retaliate against current and former employees who raise safety concerns. Additionally, they call AI companies to have a verifiably anonymous process for current and former employees to raise risk-related concerns to the appropriate bodies. The authors hope the proposed principles will bring transparency to the AI industry, encourage more people to speak out and make AI companies accountable to the public.
OpenAI Offers a Peek Inside the Guts of ChatGPT
OpenAI has released a new paper that introduces a method to identify how AI models store and process certain concepts, which could help in understanding and mitigating potential harmful behaviours of AI systems. The research was conducted by the now-disbanded Superalignment team at OpenAI, which is why the paper lists Ilya Sutskever and Jan Leike as its authors (both of them have recently left OpenAI). The technique proposed involves using an additional machine learning model to peer inside large language models like GPT-4, making their internal workings more transparent. This could aid in controlling unwanted behaviours and tuning the models for specific purposes.
▶️ I Built a CoPilot+ AI PC (without Windows) (12:49)
Raspberry Pi joins the AI party with the release of Raspberry Pi AI Kit which uses the Hailo AI accelerator to give the Raspberry Pi 13 TOPS of inferencing performance. Although the AI Kit is not powerful enough to run large language models with an acceptable level of performance, it is perfect for running image recognition, pose estimation, or other less demanding AI software on the edge or in robotics. Jeff Geerling checks what the AI Kit offers, and in a follow-up video, he pushes the idea of an AI-accelerated Raspberry Pi to the limit by combining 8 neural processors to build a 55 TOPS tiny monster.
AI researchers build ‘future self’ chatbot to inspire wise life choices
Researchers from MIT have built Future You, an AI-powered chatbot that simulates your 60-year-old future self (it even uses digitally aged photos as profile pictures) and gives wise advice about career, life, and health. “The goal is to promote long-term thinking and behaviour change,” said Pat Pataranutaporn, who works on the Future You project at MIT’s Media Lab. “This could motivate people to make wiser choices in the present that optimise for their long-term wellbeing and life outcomes.” According to the researchers, 344 volunteers have reported feeling less anxious and more connected to their future selves after having a conversation with the chatbot.
If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?
🤖 Robotics
1X AI Update: Voice Commands & Chaining Tasks
1X, a Norwegian humanoid robotics startup, showcases new capabilities of their humanoid robots in their AI update, such as taking voice commands and chaining tasks to help clean the office. The company says that the video above does not contain teleoperation, computer graphics, cuts, video speedups, or scripted trajectory playback.
Chinese automaker to use UBTECH humanoid
Chinese carmaker Dongfeng Liuzhou Motor has become the next auto company, after Mercedes and BMW, to use humanoid robots in its factory. The company has agreed to deploy an undisclosed number of UBTECH’s Walker S humanoid robots, which will perform various tasks in the car manufacturing process. Dongfeng highlighted that the humanoids will work alongside traditional automation equipment, allowing them to handle complex scenarios in unmanned production.
The Age of the Drone Police Is Here
In October 2018, Chula Vista, a city near San Diego and at the border with Mexico, launched the Drone as First Responder (DFR) program, the first of its kind in the US. The Chula Vista Police Department says DFR reduced unnecessary police contacts, decreased response times, and saved lives. However, the picture is more complicated. WIRED investigators analyzed 22 million coordinates from police drone flight paths and, in this article, paint a complex picture of the trade-offs between public safety and privacy.
▶️ Dieter Fox: Where's RobotGPT? (1:11:47)
Robotics and AI researchers have been using generative AI models for some time now to equip robots with open-world visual understanding and reasoning capabilities. However, we are still far from achieving a "RobotGPT" moment. A crucial reason for this problem is the lack of data suitable to train powerful, general models for robot decision-making and control. In this lecture, Dieter Fox, a professor at the University of Washington and Senior Director of Robotics Research at Nvidia, discusses the progress the robotics and AI community has made in generating large datasets for training robots in simulated environments. He also demonstrates a robust sim-to-real transfer of skills trained in simulation and presents a path toward generating large-scale demonstration sets that could help train robust, open-world robot manipulation models.
🧬 Biotechnology
Scaling CRISPR Cures
Unlike traditional small molecule drug discovery, which involves screening vast libraries of chemicals, CRISPR-based treatments can be rapidly developed and customized using a standard set of components, with only minor changes to the guide RNA needed to target different diseases. CRISPR could be a "platform for cures" that can be scaled to treat a wide range of genetic diseases more efficiently than traditional methods. However, there are several challenges preventing this from happening. This article by
Making Bioprocessing More Sustainable
The bioprocessing industry holds potential for significant contributions to the United Nations' sustainable development goals. However, the industry's environmental impact is largely underexplored, with studies covering only about 0.2% of pharmaceuticals and showing inconsistent results across different products, as this article reports. The article also gives some ideas on how to improve sustainability in the industry.
💡Tangents
Origins of the Lab Mouse
In this article published at
Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.
Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!
My DMs are open to all subscribers. Feel free to drop me a message, share feedback, or just say "hi!"