Biden updates AI chip export rules to China, and OpenAI, Adobe and Google headline a big month for Generative AI

Sign up
Updates

Policy.ai has temporarily moved to a once-a-month schedule. Want to make sure you don’t miss any of CSET’s expert analysis? Update your preferences!

Worth Knowing

OpenAI, Adobe and Google Headline Another Big Month for Generative AI: After a summer of lagging user growth for tools like OpenAI’s ChatGPT, some observers speculated that the generative AI boom was beginning to fizzle. But a new wave of product updates and investment news has it looking like the party’s not over yet:
  • OpenAI debuted a handful of eagerly anticipated upgrades to two of its most popular tools: ChatGPT and DALL-E. Paying ChatGPT users can now speak aloud with the app, which responds using one of several synthetic voices. The tool can also now interpret uploaded images, a multimodal capability that OpenAI showed off when it debuted GPT-4 earlier this year but waited to roll out to customers. The company also announced DALL-E 3, an update to the popular text-to-image tool released in 2021. The update is also accessible to paying customers through ChatGPT and through Microsoft’s Bing chat.  
  • Google connected its “Bard” chatbot to its suite of tools, meaning users can have Bard engage with their data from Gmail, Docs, Drive, and other Google services. The tech giant debuted Bard earlier this year in response to the perceived threat posed by ChatGPT, but Google’s product has yet to catch on like its rival. Tighter integration with Google’s immensely popular tools could be what it takes to turn the tide.
  • Adobe announced an update to its own AI-powered image generator, “Firefly.” While the update itself is not particularly game-changing — comparisons posted online show it performs similarly to competitors like DALL-E and Midjourney — Adobe’s tool could have a built-in advantage over those other systems: the company says Firefly was trained exclusively on licensed and public domain content. With the environment around generative AI training becoming increasingly litigious, Adobe’s safeguards could make it a go-to for more risk-averse companies and individual content creators.
The steady drumbeat of announcements and integrations is one thing, but we also gained some insight into one of the key questions surrounding the burgeoning industry: is generative AI making any money? According to a report from The Information, OpenAI is set to pull in more than $1 billion in revenue over the next year. It’s not clear whether that’s enough to break even — the company lost more than $500 million last year — but the revenue numbers are nevertheless impressive for a company that only debuted its flagship product last November. Investors, for their part, appear bullish: last month, Amazon announced plans to invest $4 billion in OpenAI rival Anthropic.
Tech Giants Explore Custom Chips — Should Nvidia Worry?: OpenAI is considering a plan to design its own AI chips, according to a recent Reuters report. While the plan is still in its infancy, Reuters says the company has identified a potential chip firm to acquire. OpenAI is not alone in trying to enter the hardware space — major AI developers like Google, Meta, and Amazon have worked on their own custom chips, to varying degrees of success. According to another report by The Information, Microsoft — which has invested $13 billion in OpenAI and integrated its models across its services — has begun testing its own custom chips. A broader shift toward in-house chip design, as well as new AI-focused chips from Nvidia rivals like AMD and Intel, could be bad news for Nvidia, which has ridden its dominant position as the AI chip developer of choice to a trillion-dollar valuation. But building a serious competitor to Nvidia is about more than well-designed silicon; Nvidia’s tightly integrated software and hardware mean that adopting a new chip supplier is more difficult than pulling out an Nvidia GPU and plugging in a competitor’s. Nevertheless, Nvidia doesn’t seem content to rest on its laurels — industry observer SemiAnalysis speculates that the company accelerated plans for next generation AI GPUs in response to the threat posed by the likes of Google, Amazon, Microsoft, AMD, and Intel.
Government Updates

White House Updates Rules on Chips and Chipmaking Exports to China: Last week, the Biden administration announced “targeted updates” to tighten rules restricting the export of powerful AI chips and chipmaking equipment to China. As with the controls announced on October 7th of last year, the goal of the updates is to cut off China’s access to the chips needed for high-end AI development and other compute-dependent military applications, according to comments from Commerce Secretary Gina Raimondo. While not a sea change, the updates nevertheless aim to more effectively achieve national security objectives. Changes include:
  • Updating the metrics that define controlled chips by replacing networking speed with performance density. After the original October 7th controls, chip firms such as Nvidia developed chips for export to the Chinese market (in Nvidia’s case, the A800 and H800) that were just below the networking threshold set by the rules but still capable of running high-end AI workloads. According to Nvidia, the updated rules now cover the A800 and H800, but also the less powerful L40 and L40S enterprise GPUs, as well as the company’s highest-end consumer gaming GPU, the RTX 4090.
  • Adding countries to the list of restricted destinations for high-end chips. In addition to China and Macau, the updated rules place restrictions on more than 40 countries due to national security concerns, including the risk of diversion to China. 
  • Extending the list of restricted semiconductor manufacturing equipment exports to cover a variety of tools, including an additional ASML deep ultraviolet lithography system that can be used to make advanced (but not bleeding-edge) chips. The restricted destinations for this new list of tools has been expanded to include 22 additional countries beyond China.
While some provisions of the updated rules went into effect immediately, the bulk of the rules are set to go into effect next month. This probably isn’t the last time they’ll be revised, either — Raimondo said Commerce plans to update the rules “at least annually.”

AI in the Senate — Schumer’s Forums, Thune & Klobuchar’s “Light Touch” and a Blumenthal-Hawley Framework: Earlier this week, lawmakers and tech and policy leaders convened for the second in a series of nine “AI insight forums” organized by Senate Majority Leader Schumer as part of his plan to develop “comprehensive legislation” to regulate AI. But it appears not everyone is on board with Schumer’s roadmap — a number of senators have said that the Senate needs to move faster on AI regulation and argued that Schumer’s educational push is slowing down the legislative process. Senators Thune and Klobuchar reportedly plan to introduce a “light touch” AI bill that would require AI developers to test and self-certify the safety of their systems. Senators Blumenthal and Hawley have also outlined a legislative framework that would require licenses for high-risk or sophisticated general purpose models and require certain safety and transparency measures from private AI developers. Klobuchar and Thune, the Senate Minority Whip, seem to have deviated from Schumer on both speed and substance — Thune told Politico that his planned bill is meant to take a different tack from the “heavy-handed regulatory approaches” likely to be pursued by Schumer and others. Thune and Klobuchar said they would likely introduce their bill this month. But another important regulatory step could come even sooner — according to The Washington Post, the White House plans to release an executive order before next week’s UK AI Summit that would subject AI models to assessment before they can be eligible for federal use.

In Translation
CSET’s translations of significant foreign language documents on AI

Chinese Integrated Circuit Subsidy Notice: Project Application Guide for the Shenzhen Science and Technology Innovation Commission’s 2024 Annual Integrated Circuit Special Project Subsidy Program. This notice — one of many local Chinese semiconductor industry incentive policies — describes how integrated circuit design and manufacturing companies can apply for local government subsidies in Shenzhen, a major tech hub in southern China. Shenzhen’s program subsidizes projects related to Integrated Circuit (IC) tape-out the most generously, with lesser amounts available for IC design and electronic design automation (EDA) software development. The notice threatens applicants with blacklisting or worse for submitting fraudulent applications, suggesting that such conduct on the part of applicants is not unheard of.

If you have a foreign-language document related to security and emerging technologies that you’d like translated into English, CSET may be able to help! Click here for details.

Job Openings

We’re hiring! Please apply or share the roles below with candidates in your network:
  • Research Fellow – CyberAI: We are currently seeking candidates who are passionate about exploring topics at the intersection of AI and cybersecurity. As a CyberAI Research Fellow, you will play a pivotal role in assessing how AI techniques can enhance cybersecurity, mitigate emerging threats, and shape future cyber operations. We value candidates who can bridge the gap between technical and non-technical audiences, making complex concepts accessible. If you possess strong analytical skills, practical experience in machine learning or cybersecurity, and a deep understanding of AI’s failure modes and potential threats, we want to hear from you. Apply by Monday, November 13th
  • Senior or Research Fellow – Compete LOR Lead: We are currently seeking candidates to lead our Compete Line of Research as a Senior or Research Fellow. This Fellow will play a pivotal role in leading and coordinating our Compete LOR efforts, by shaping priorities, devising research strategies, and overseeing research execution and report production. We are looking for individuals who possess a profound understanding of technological innovation’s pivotal role in U.S. national power, with expertise in areas such as economic security, trade and investment controls, trade regulation, and antitrust laws. Apply by Monday, November 13th
Please bookmark our careers page to stay up to date on all active job postings. You can also subscribe to receive job announcements by updating your subscription preferences in the footer of this email.

What’s New at CSET

REPORTS

PUBLICATIONS AND TRACKERS

EMERGING TECHNOLOGY OBSERVATORY

EVENT RECAPS

  • On September 13, CSET’s Margarita Konaev, Emelia Probasco, Jack Corrigan and Ali Crawford discussed ways the United States can promote innovation to maintain its competitive advantage in emerging technologies.

IN THE NEWS

What We’re Reading

Paper: Open (For Business): Big Tech, Concentrated Power, and the Political Economy of Open AI, David Gray Widder, Sarah West and Meredith Whittaker (August 2023)

Discussion Paper: Frontier AI: Capabilities and Risks — UK AI Summit Discussion Paper, UK Department for Science, Innovation and Technology (October 2023)

Information Sheet: Contextualizing Deepfake Threats to Organizations, the National Security Agency, the Federal Bureau of Investigation and the Cybersecurity and Infrastructure Security Agency (September 2023)

Upcoming Events


What else is going on? Suggest stories, documents to translate & upcoming events here.