Lights, Cameras, Action!
The Trump administration has released its “Winning the AI Race: America's AI Action Plan”, a design for winning the AI race and becoming the dominant global AI power. As we noted last week, rather than focusing on developing a rational strategy that includes insuring the health and safety of the US population and the planet, the focus of this strategy, not surprisingly, is to win, win, win and not worry about the consequences (until after we win).
There are over 90 ‘points’ in the document, although many of these are recommendations and goals rather than legally binding actions. At the same time, three EOs that are legally binding were also signed, one of which we have already reviewed (Promoting the Export of the American AI Technology Stack). The other two are, “Accelerating Federal Permitting of Data Center Infrastructure” and “Preventing Woke AI in the Federal Government”. The former removes any hope of climate change restrictions when it comes to data centers, and the latter is another anti-DEI tirade specific to AI and data centers. Intermixed are plans for government financing for large data centers, although much is boiler plate and a few small tidbits for the former tree huggers that now want to avoid a global power shortage, climate change, or the enslavement of the human race by AI overlords!.
Rather than walk through all 90+ recommendations in the plan, we looked at it according to 14 government agencies that were singled out in the plan. However, unless the agency was specifically mentioned as a focus in the three additional EOs, the plan does not mandate any of these actions, it only recommends them. So for those actions in the plan that are not part of the EOs life can go on as it has thus far without assigned change. In other words, as it was before the action plan. That said, looking at each agency against the EOs, anything underlined and in italics becomes mandated, anything not is a suggestion.
When all is said and done, the ‘Action Plan’ is nothing more than a reiteration of the three executive orders that preceded it, as it carries no weight other than what is mandated in the three EOs (underlined here). Simply put they are:
The only reference to anything outside of direct promotion of the EO mandates and the AI race was the mention of open source as a way to promote the development of Ai applications, an important point that is a bit out of character for the current administration’s policy of capital formation at all costs. Open source models allow developers, especially smaller developers, to utilize models that would typically be available only to large institutions and enterprise customers, and as we have said in the past, the race is not to build the best model, it is to build the most practical applications.
The average smartphone user likes the idea that AI is helping to take better pictures but could do without it if necessary. However if someone came up with an AI based application that could sort through your e-mails, put them in priority order, write responses to those that needed them (with your approval), make notations to your calendar, read all of your texts and note important events and personal nuances, and make sure you were aware that the #2 subway was stalled at 14th street (police investigation) and an alternative might be necessary, and Microsoft (MSFT) did not own it (free is good), then there would be real value in AI to the average consumer. Other than that its all promotion.
Here are the ‘Plan’ details:.
I. Office of Management and Budget (OMB)
There are over 90 ‘points’ in the document, although many of these are recommendations and goals rather than legally binding actions. At the same time, three EOs that are legally binding were also signed, one of which we have already reviewed (Promoting the Export of the American AI Technology Stack). The other two are, “Accelerating Federal Permitting of Data Center Infrastructure” and “Preventing Woke AI in the Federal Government”. The former removes any hope of climate change restrictions when it comes to data centers, and the latter is another anti-DEI tirade specific to AI and data centers. Intermixed are plans for government financing for large data centers, although much is boiler plate and a few small tidbits for the former tree huggers that now want to avoid a global power shortage, climate change, or the enslavement of the human race by AI overlords!.
Rather than walk through all 90+ recommendations in the plan, we looked at it according to 14 government agencies that were singled out in the plan. However, unless the agency was specifically mentioned as a focus in the three additional EOs, the plan does not mandate any of these actions, it only recommends them. So for those actions in the plan that are not part of the EOs life can go on as it has thus far without assigned change. In other words, as it was before the action plan. That said, looking at each agency against the EOs, anything underlined and in italics becomes mandated, anything not is a suggestion.
When all is said and done, the ‘Action Plan’ is nothing more than a reiteration of the three executive orders that preceded it, as it carries no weight other than what is mandated in the three EOs (underlined here). Simply put they are:
- Prevent woke AI in the Federal government
- Accelerating Federal permitting of data center infrastructure
- Promoting the export of the American AI Technology stack.
The only reference to anything outside of direct promotion of the EO mandates and the AI race was the mention of open source as a way to promote the development of Ai applications, an important point that is a bit out of character for the current administration’s policy of capital formation at all costs. Open source models allow developers, especially smaller developers, to utilize models that would typically be available only to large institutions and enterprise customers, and as we have said in the past, the race is not to build the best model, it is to build the most practical applications.
The average smartphone user likes the idea that AI is helping to take better pictures but could do without it if necessary. However if someone came up with an AI based application that could sort through your e-mails, put them in priority order, write responses to those that needed them (with your approval), make notations to your calendar, read all of your texts and note important events and personal nuances, and make sure you were aware that the #2 subway was stalled at 14th street (police investigation) and an alternative might be necessary, and Microsoft (MSFT) did not own it (free is good), then there would be real value in AI to the average consumer. Other than that its all promotion.
Here are the ‘Plan’ details:.
I. Office of Management and Budget (OMB)
- Deregulatory Actions:
- Work with all Federal agencies to identify, revise, or repeal regulations, rules, memoranda, administrative orders, guidance documents, policy statements, and interagency agreements that hinder AI innovation and adoption.
- Work with Federal agencies that have AI-related discretionary funding programs to ensure, consistent with applicable law, that they consider a state's AI regulatory climate when making funding decisions and limit funding if the state's AI regulatory regimes may hinder the effectiveness of the funds.
- Issue guidance to agencies regarding "unbiased AI principles" for federal procurement of LLMs.
- National Institute of Standards and Technology (NIST):
- Revise the NIST AI Risk Management Framework to eliminate references to misinformation, Diversity, Equity, and Inclusion (DEI), and climate change.
- Lead NAIRR (National AI Research Resource) pilot efforts to give startups and researchers better access to compute, data, and software (in coordination with OSTP and NSF).
- Convene public and private sector stakeholders to accelerate the development and adoption of national standards for AI systems (e.g., in healthcare, energy, agriculture).
- Publish guidelines and resources for federal agencies to conduct their own evaluations of AI systems.
- Develop a formal guideline and companion voluntary forensic benchmark for deepfake evaluation (building on Guardians of Forensic Evidence program).
- Convene the NIST AI Consortium to develop scalable, interoperable measurement techniques and metrics.
- Support theoretical, computational, and experimental research to foster new AI breakthroughs.
- National Telecommunications and Information Administration (NTIA):
- Convene stakeholders to drive adoption of open-source and open-weight models by small and medium-sized businesses.
- Economic Development Administration (EDA) / Other DOC bureaus:
- Launch an initiative to provide financial support (loans, grants, tax incentives, off-take agreements) for qualifying data center, semiconductor manufacturing, and energy infrastructure projects.
- Continue CHIPS program funding and simplify administrative processes.
- Partner with industry to deliver secure, full-stack AI export packages to allied nations.
- Establish an American AI Exports Program.
- Implement new export controls on semiconductor manufacturing subsystems and expand US oversight on foreign-made items that rely on US technology.
- Workforce Development:
- Prioritize AI skill development as a core objective of education and workforce funding streams.
- Establish an "AI Workforce Research Hub" to evaluate AI's impact on the labor market.
- Create a national initiative identifying high-priority occupations critical to AI infrastructure (e.g., electricians, HVAC technicians) and partner with state and local governments to support industry-driven training programs.
- Expand Registered Apprenticeships for AI-critical occupations and update career and technical education programs.
- Create early career exposure programs and pre-apprenticeships for middle and high school students.
- AI Adoption and Research:
- Transform senior military colleges into specialized hubs of AI R&D.
- Integrate AI-specific curriculum across various academic majors to foster military AI expertise.
- Refine its Responsible AI and Generative AI Frameworks, Roadmaps, and Toolkits.
- Prioritize and invest in the construction of secure data centers for national security applications.
- Prioritize agreements with cloud service providers and computing infrastructure operators for priority access to computing resources in a national emergency.
- Defense Advanced Research Projects Agency (DARPA):
- Launch a technology development program to improve AI interpretability and control.
- Energy Infrastructure & Scientific Computing:
- Expand the number of agencies participating in the Department of Energy's PermitAI project to expedite permitting for data centers.
- Support automated, cloud-enabled labs across various scientific fields in collaboration with National Laboratories.
- Promote dispatchable, next-generation energy sources (enhanced geothermal, nuclear fission, fusion).
- Cybersecurity & Incident Response:
- Establish an AI-ISAC (Information Sharing and Analysis Center) to circulate AI security threat intelligence.
- Develop guidance for the private sector regarding remediating and responding to AI-specific vulnerabilities and threats.
- Promote the development and incorporation of AI Incident Response actions into existing incident response doctrine and best practices.
- (With NIST) Include AI in their standards and guidance related to incident response.
- Policy Coordination & Deregulation:
- Lead a Request for Information from businesses and the public about current Federal regulations that hinder AI innovation and adoption.
- Coordinate with NIST and NSF on NAIRR pilot efforts.
- Participate in CAISI intelligence collection (with NSC, IC, DOE).
- International Diplomacy & Exports:
- Partner with industry through the Commerce and State Departments to deliver secure, full-stack AI export packages to allied nations.
- Actively counter China's growing presence in international AI governance bodies.
- Strengthen global alliances to maintain trust and promote responsible AI practices.
- Tax Incentives:
- Issue guidance clarifying that many AI literacy and skill programs may qualify as eligible education assistance under the Internal Revenue Code, allowing employers to offer tax-free reimbursement.
- Legal System & Synthetic Media:
- Consider developing a formal guideline and companion voluntary forensic benchmark for deepfake evaluation to equip the legal system.
- Combat industrial espionage and theft of AI technology.
- Research & NAIRR:
- Participate in NAIRR pilot efforts (with NIST, OSTP).
- Support theoretical, computational, and experimental research.
- Chief Artificial Intelligence Officer Council (CAIOC): Formalize as the primary venue for interagency coordination.
- Cross-Agency Directives:
- Direct federal agencies to identify and eliminate rules that may slow the deployment of AI technologies.
- Direct federal agencies to assess and expand their use of AI to improve efficiency and effectiveness.
- Direct agencies to build a pipeline of AI-skilled federal employees.
- Direct federal agencies to identify and make available appropriate federal lands for data center construction.
- AI Safety Evaluation:
- Prioritize, collect, and distribute intelligence on foreign frontier AI projects with national security implications (in collaboration with DOE, NSC, IC, and OSTP).
- Evaluate frontier models from China for alignment with Chinese Communist Party (CCP) censorship and talking points.
- Hold biannual CAISI-led meetings for federal agencies and researchers to share AI evaluation best practices.
- Federal Communications Commission (FCC): (Potentially) Review state AI regulations that interfere with its mandate.
- Federal Trade Commission (FTC): Review and reassess ongoing AI-related investigations and final orders that burden AI innovation.
- Securities and Exchange Commission (SEC) and Food and Drug Administration (FDA): Encouraged to use regulatory sandboxes.
- National Science and Technology Council's Machine Learning and AI Subcommittee: Directed to develop minimum data quality standards for scientific disciplines.
RSS Feed