Earlier this week, the President’s Council of Advisors on Science and Technology (PCAST) released a report outlining what it believes must happen for the U.S. to advance “industries of the future.” Several of the committee’s suggestions touched on the field of AI as it relates to federal, state, and private-sector partnerships, as well as departmental budgetary considerations. In particular, the report recommends that the U.S. grow nondefense federal investments in AI by 10 times over the next 10 years and for the federal government to create national AI “testbeds,” expanding the National Science Foundation’s (NSF) AI Institutes with at least one AI Institute in each state and creating “National AI Consortia” to share capabilities, data, and resources.
Loosely, PCAST — which lives in the Office of Science and Technology — provides advice to the president on science and technology policy. (Its 12 members from academia and private industry met for the third time this week under the Trump Administration.) In the report, the committee argues the U.S. will need to boost AI R&D investments from $1 billion a year in 2020 to $10 billion a year by 2030 in order to remain competitive. PCAST asserts this would enable the NSF — which requested $487 million for AI in 2020 — to make at least 1,000 awards to individual investigators “without any loss of quality.”
PCAST also recommends driving opportunities for AI education and training, in part by:
- Securing pledges to scale investments on training and education of the U.S. workforce in AI
- Developing AI curricula and performance metrics at K-12 through postgraduate levels and for certificate and professional programs
- Training a highly skilled AI workforce at secondary schools and universities
- Creating incentives, recruitment, and retention programs for AI faculty at universities
- Increasing NSF and Department of Education investments in AI educators, scientists, and technologists at all levels
Education and immigration
Laments over the AI talent shortage in the U.S. have become a familiar refrain from private industry. According to a report by Chinese technology company Tencent, there are about 300,000 AI professionals worldwide but “millions” of roles available. In 2018, Element AI estimated that of the 22,000 Ph.D.-educated researchers globally working on AI development and research, only 25% are “well-versed enough in the technology to work with teams to take it from research to application.” And a Gartner survey found that 54% of chief information officers view this skills gap as the biggest challenge facing their organization.
While higher education enrollment in AI-relevant fields like computer science has risen rapidly in recent years, few colleges have been able to meet student demand due to a lack of staffing. There’s evidence to suggest the number of instructors is failing to keep pace with demand due to private sector poaching; from 2006 to 2014, the proportion of AI publications with a corporate-affiliated author increased from about 0% to 40%, reflecting the growing movement of researchers from academia to corporations.
Europe tellingly overtook the world in scholarly output related to AI last year, according to a report by Elsevier. China, whose “AI Innovation Action Plan for Colleges and Universities” called for the establishment of 50 new AI institutions by 2020, is expected to overtake the EU within the next four years if current trends continue.
PCAST suggests a remedy in stronger collaboration with “key U.S. allies,” including formal international partnerships in AI research and development. Unfortunately, the Trump Administration’s overtures make this perhaps the least realistic of the committee’s goals. In a bid to pressure schools to reopen during the pandemic, U.S. Immigration and Customs Enforcement recently said it would force out international students who don’t attend in-person classes. And in June, the Trump Administration imposed a ban on entry into the U.S. for workers on certain visas — including for high-skilled H-1B visa holders, an estimated 35% of whom have an AI-related degree — through the end of the year.
Even before the new visa restrictions, immigration obstacles had begun to hurt AI activity in the U.S. Companies like Facebook, Microsoft, Google, Amazon, and Intel established AI centers in other countries in pursuit of local talent; Apple director of machine learning Ian Goodfellow called the U.S.’s immigration policy “one of the largest bottlenecks to [the AI community’s] collective research productivity over the last few years.”
“It is quite clear that AI is going to touch every area of science,” director of IBM Research PCAST member Dario Gil said during a virtual press briefing on Wednesday. “Over the last decade, powered by exponential growth in computing power, and ever increasing availability of data, technological breakthroughs in AI are enabling intelligent systems to take on increasingly sophisticated tasks and augmenting human capabilities in a new and profound way … It is important to create a virtuous cycle aimed at the innovation infrastructure itself to continuously accelerate R&D in AI.”
New funding
Ten billion dollars within the next 10 years might sound ambitious, but it’s in line with the budgets already approved by countries with national AI research initiatives. For instance, Canada’s Pan-Canadian Artificial Intelligence Strategy is a five-year, $94 million (CAD $125 million) plan to invest in AI research and talent, complementing the government’s investments of nearly $173 million (CAD $230 million) and $45 million (CAD $230 million) in Scale.AI, a business-led consortium. The EU Commission has committed to increasing investment in AI from $565 million (€500 million) in 2017 to $1.69 billion (€1.5 billion) by the end of 2020. France recently took the wraps off a $1.69 billion (€1.5 billion) initiative aimed at transforming the country into a “global leader” in AI research and training. And in 2018, South Korea unveiled a multiyear, $1.95 billion (KRW 2.2 trillion) effort to strengthen its R&D in AI, with the goal of establishing six AI-focused graduate schools by 2022 and training 5,000 AI specialists.
If anything, PCAST’s recommendation falls on the conservative side of the spectrum. When Michael Kratsios, the U.S. chief technology officer, revealed last September that U.S. government agencies requested nearly $1 billion in nondefense AI research spending for the fiscal year ending in September 2020, representatives from Intel, Nvidia, and IEEE said the U.S. would need to invest even more in AI. (The White House earlier this year proposed setting aside an additional $1 billion for a total of $2 billion by 2022.) Separately, national security think tank Center for a New American Security called for federal spending on high-risk/high-reward AI research to increase to $25 billion by 2025 to avoid “brain drain.”
Beyond funding, PCAST advocates for the creation of AI investment pledges to support universities and for tasking the National Institute of Standards and Technology and the National Institutes of Health with curating, managing, and disseminating “AI-ready” data sets. These efforts could lead to cheaper compute infrastructure for research and new open source AI frameworks and tools, the subcommittee said, and they’d complement the state-by-state AI Institutes’ work on things like agriculture or AI for manufacturing and “AI for social good.”
“We cannot emphasize enough how critical it is to get data AI-ready. Let’s remember that 80% of the effort of any AI project is typically spent on the data curation and preparation,” Gil continued. “That is why we recommend expanding the ongoing NSF-based programs to establish national AI research centers and infrastructure with sustained long-term funding to enable cross-cutting research and technology transitions … In recent months, during the COVID-19 crisis, AI has demonstrated critical capabilities as well as important potential for the future.”
Author: Kyle Wiggers.
Source: Venturebeat
