US Expands AI Chip Export Ban - Weekly Roundup - Issue #438
This week - Spot can now speak and has a personality; NO FAKES Act; Amazon Prime Air is expanding; Baidu reveals Ernie 4.0; and more!
A thriving AI industry stands on three main pillars: a skilled workforce to design advanced AI systems, vast amounts of data to train algorithms, and enormous computing power to train and run AI systems. As the tension between the US and China grows, the US government tries to curb China’s AI capabilities by restricting its access to cutting-edge GPUs.
The first ban on advanced AI chip export was announced in September 2022. The original rules took into account the computing power of the chips and how fast they can communicate with each other. This affected many of the market's most powerful GPUs, including heavily sought-after Nvidia’s A100 and H100 chips. In reaction, Nvidia designed and released new chips, the A800 and H800, to comply with those export rules for the Chinese market.
The chip bans were originally targeted at China’s military use, but the ban also affects Chinese AI startups and companies. Many of them rushed to stockpile Nvidia’s A100 and H100 before the bans went into effect, no matter the cost or the source. Chinese tech giants like Alibaba, Baidu, ByteDance, and Tencent are reported by the Financial Times to have collectively placed orders worth $5 billion for A800 chips for this and the upcoming year.
But now the rules have been changed. On October 17th, the Biden administration announced an update on exporting advanced AI chips to China and other countries. Starting from November 16, 2023, Nvidia, Intel, AMD and other US semiconductor manufacturers will be unable to ship their most powerful products and modules for AI and high-performance computing to China, Saudi Arabia, the United Arab Emirates, and Vietnam without an export license from the US Department of Commerce's Bureau of Industry and Security (BIS).
Recent reports suggest that the updated export regulations might take effect as early as October 30th, two weeks sooner than originally planned.
The new rules will focus on computing power alone, which will encompass more chips than the original ban. In fact, the restrictions are so severe that they will include not only top computing products - such as Nvidia’s A100, A800, H100, H800, L40 and L40S - but also top-end gaming graphics cards, like Nvidia’s GeForce RTX 4090. Primarily designed for high-end gamers and professionals, RTX 4090 are also used by AI researchers as a cheaper and more accessible alternative to A100 or H100 chips.
In a recent statement, Nvidia downplayed any immediate financial repercussions from these developments, saying, "Given the strength of demand for the Company’s products worldwide, the Company does not anticipate that the accelerated timing of the licensing requirements will have a near-term meaningful impact on its financial results."
A spokesperson for the Chinese embassy said, quoted by Reuters, that it "firmly opposes" the new restrictions, adding that "arbitrarily placing curbs or forcibly seeking decoupling to serve (a) political agenda violates the principles of market economy and fair competition (and) undermines the international economic and trading order."
It's important to clarify that the new rules do not impose a blanket export ban on all chips to China. Officials at the BIS said they are open to hearing the semiconductor industry's input for finding ways to keep sending AI chips to China for small and medium-sized systems and to find a "tamperproof" way to keep systems that might contain up to 256 AI chips from being strung together into a supercomputer. "This approach could constrain (controlled AI chips) from being used to train large dual-use AI foundation models with capabilities of concern, while allowing AI training capabilities at a small or medium scale," the BIS wrote, as reported by Reuters.
If you enjoy this post, please click the ❤️ button or share it.
Becoming a paid subscriber is the best way to support the newsletter.
If you enjoy and find value in my writing, please hit the like button and share your thoughts in the comments. Share the newsletter with someone who will enjoy it, too.
You can also buy me a coffee if you enjoy my work.
🦾 More than a human
Above-Elbow Bionic Arm Can Control Every Finger
Researchers from Sweden have developed a robotic hand that interfaces directly with the user's nervous and muscular systems. Specifically designed for individuals with above-elbow amputations, this prosthetic hand connects to nerve bundles and integrates with new muscles. The prosthetic has been put to the test in real-life scenarios: a patient has been using it for over three years, assisting him in daily activities like grasping objects and pouring drinks.
🧠 Artificial Intelligence
Senate Legislation Would Outlaw Unauthorized AI-Generated Likenesses; SAG-AFTRA Lauds “No Fakes Act”
Four US senators have introduced the NO FAKES Act, a draft bill designed to protect actors, singers, and other artists from unauthorized AI-generated replicas of their voices and likenesses. If passed, individuals or platforms producing or hosting such unauthorized digital replicas could be held liable. The concerns about studios using AI to replace actors were one of the reasons why screenwriters and actors went on strike in May.
China's Baidu unveils new Ernie AI version to rival GPT-4
Chinese tech giant Baidu has unveiled its latest generative AI model, Ernie 4.0, which it claims is on par with OpenAI's GPT-4. Baidu's CEO, Robin Li, demonstrated the model's memory capabilities by having it write a martial arts novel in real time and produce advertising materials. However, the launch failed to greatly impress analysts, many of whom noted a lack of significant improvements over the previous version.
OpenAI: Frontier risk and preparedness
As part of their mission to create safe artificial general intelligence (AGI), OpenAI has announced the formation of a new team called Preparedness. This team will oversee capability assessment, evaluations, and internal red teaming for frontier models. Their primary areas of focus include individualized persuasion, cybersecurity, threats ranging from chemical to nuclear, and autonomous replication and adaptation. Additionally, they will manage the Risk-Informed Development Policy (RDP), which will outline their approach to developing, monitoring, and overseeing frontier model capabilities. OpenAI is also launching the AI Preparedness Challenge to identify less obvious areas of concern and prevent catastrophic misuse of AI. If you have a unique or intriguing story about AI going wrong, you can submit it to OpenAI for a chance to win prizes and maybe even to join the Preparedness team.
‘Here is the news. You can’t stop us’: AI anchor Zae-In grants us an interview
In this article, The Guardian speaks with Zae-In, a South Korean news anchor. However, Zae-In is not a typical anchor; she is an AI presenter, one of many that have emerged in recent years in countries such as China, India, Taiwan, Greece, and Kuwait. These AI presenters are appealing—they are attractive, ageless, can work 24/7, and don't cause any problems. Yet, the rise of AI anchors has been met with mixed reactions. There are concerns regarding trust, especially when AI newsreaders are used for spreading propaganda.
🤖 Robotics
Amazon’s Prime Air is coming to a new US city
Prime Air - Amazon’s drone delivery service - is expanding to Italy and the UK. Additionally, the service is going to be available in the third city in the US. Amazon did not specify which cities the service is coming to, saying names of chosen cities will be revealed in the coming months.
▶️ Making Chat (ro)Bots (8:27)
Boston Dynamics’ Spot has learned a new skill - it can now talk. Not only that, but it can also assume any personality, from a fancy butler to an excited tour guide to "Josh." Currently, the talking Spot is the result of a hackathon and is in the early research stages. However, it is possible that soon we could command robots the same way that we speak with other humans. The use of generative AI chatbots can also expand the capabilities of the robot by providing additional context about the environment or the task at hand.
🧬 Biotechnology
Thousands of programmable DNA-cutters found in algae, snails, and other organisms
Researchers from MIT have discovered thousands of programmable DNA-cutting enzymes, known as Fanzors, in various species such as snails, algae, and amoebas. Similar to CRISPR, Fanzors are RNA-guided enzymes that can be tailored to cut DNA at specific locations, potentially allowing for more precise gene-editing tools.
3D-Printed Living Materials Glow Under Mechanical Stress
Inspired by the bioluminescent waves seen during red tide events at San Diego's beaches, researchers from the University of California, San Diego (UCSD) have created a living composite material that glows under mechanical stress. They hope these living composites can be utilised in robotics to construct biohybrid robots or in biomedical applications like in vivo light sources for targeted drug release, photothermal therapy, and photodynamic therapy.
Thanks for reading. If you enjoy 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!
You can follow Humanity Redefined on Twitter and on LinkedIn.
can you enable the audio? I listen to almost everything on Substack and no audio means skip