US Vice President Kamala Harris delivers a speech on Artificial Intelligence (AI) in central London, … [+] on November 1, 2023. (Photo by Daniel LEAL / AFP) (Photo by DANIEL LEAL/AFP via Getty Images)
AFP via Getty ImagesBeyond the broad brushstrokes of ethics, President Biden’s new Executive Order arms regulators with fine-tip guardrails to shape AI for the common good. Mandatory disclosures pierce the black box, requiring developers to expose critical testing data before deploying high-risk systems. Reformed immigration aims to attract brainpower for breakthroughs, but new standards demand security and fairness. Differential privacy and audits for algorithmic bias are prescribed to remedy AI’s penchant for surveillance and discrimination.
Mindmap created by Professor Marko Grobelnik – researcher in the field of Artificial Intelligence. … [+] Marko co-leads Artificial Intelligence Lab at Jozef Stefan Institute,
Professor Marko Grobelnik, researcher in the field of Artificial IntelligenceAppointing a chief AI regulator, planning for job displacement, requiring disclosures – such specifics make clear this is no vague manifesto. It pioneers precedents for oversight to forestall a dystopian AI future. The intent is unambiguous – responsible AI now takes center stage.
WASHINGTON, DC – OCTOBER 30: U.S. President Joe Biden and Vice President Kamala Harris hold an event … [+] to highlight their administration’s approach to artificial intelligence in the East Room of the White House on October 30, 2023 in Washington, DC. President Biden issued a new executive order on Monday, directing his administration to create a new chief AI officer, track companies developing the most powerful AI systems, adopt stronger privacy policies and “both deploy AI and guard against its possible bias,” creating new safety guidelines and industry standards. (Photo by Chip Somodevilla/Getty Images)
Getty ImagesCentralized AI Strategy and Mandatory Corporate Disclosures
Right up front, a defining feature of the order is the creation of a Chief AI Officer role to coordinate a unified national AI strategy across agencies. This reflects a comprehensive approach to address AI’s opportunities while mitigating specific risks that have been flagged as potentially problematic.
For instance, directives to address job losses acknowledge concerns about AI’s impact on employment and aim to ease workforce transitions. Provisions calling for safety standards seek to allay fears of national security threats from uncontrolled AI advancements. By tackling such real-world issues head-on, the order aims for responsible integration of AI.
Moreover, in the name of transparency and accountability, the EO mandates detailed disclosures requiring private companies to share results from independent third-party audits and penetration tests with federal regulators before launching any products or services incorporating advanced AI systems deemed high-risk.
By avoiding a narrow focus on specific AI technologies, the order’s language improves flexibility, which could be beneficial in the fast-evolving field of AI. The disclosures enable accountability, while the broader scope aims to future-proof oversight across various kinds of AI innovation. Together, these moves indicate efforts to balance guardrails with competitiveness.
WASHINGTON, DC – OCTOBER 30: White House Chief of Staff Jeff Zients and Senate Majority Leader … [+] Charles Schumer (D-NY) visit before an event about the Biden administration’s approach to artificial intelligence during an event in the East Room of the White House on October 30, 2023 in Washington, DC. President Joe Biden issued a new executive order on Monday, directing his administration to create a new chief AI officer, track companies developing the most powerful AI systems, adopt stronger privacy policies and “both deploy AI and guard against its possible bias,” creating new safety guidelines and industry standards. (Photo by Chip Somodevilla/Getty Images)
Getty ImagesA Flexible Alternative to Legislation
Unlike formal legislation, this executive action provides a more adaptable policy framework for shaping the evolving AI landscape. While future administrations can reverse executive orders, this order lays down a crucial early precedent that could inform future legislative efforts on AI governance.
The flexible approach enables tailoring oversight to balance guardrails with room for innovation without Congressional approval. However, formal legislation would provide more durable and democratically-validated constraints.
For now, the order allows experimenting with oversight mechanisms to address AI risks while supporting growth. It signals the private sector to align internal policies with evolving government expectations. The fluid executive direction can complement legislative efforts by pioneering frameworks that later crystallize into laws.
However, the soft law approach poses risks if compliance is inconsistent. The administration will need proactive engagement with industry and civil society to turn the order’s principles into practice. Still, agility enables adapting policies as AI capabilities advance. Striking the right balance remains a challenge.
Installing Guardrails for AI Safety and Security
A core emphasis of the order is on establishing rigorous new safety standards and testing requirements for private sector AI systems, especially those that pose significant risks such as self-driving cars, clinical decision tools, fraud detection, public infrastructure management and more.
It directs the National Institute of Standards and Technology (NIST) to develop precise methodologies and benchmarks for certifying AI systems as trustworthy based on attributes like accuracy, security, explainability and mitigation of unintended bias.
The order also establishes an AI Safety Board under the Department of Homeland Security, comprising independent experts to continuously evaluate AI used in critical infrastructure like energy, transportation, and finance by auditing internal technical documentation.
These provisions aim to install guardrails that ensure private sector AI development and deployment are safe and secure and align with ethical values. The transparency requirements make it difficult for companies to make unverified claims of trustworthiness.
However, as OctoML CEO Luis Ceze notes, “Over-indexing on regulation this early in AI can prove to be a net negative. Any government putting too many restrictions on AI will not only strangle innovation; the country in question will eventually suffer from a brain drain.”
By mandating that companies flag potentially risky models, the order promotes a culture of accountability and self-regulation. However, the implications of such disclosures remain unclear, which could create uncertainty for providers. Still, transparency enables oversight while avoiding restrictive pre-approvals that may limit innovation. This delicate balance seeks to install guardrails without stifling progress.
Mitigating AI’s Privacy Risks
The data-intensive nature of AI raises concerns about overreaching surveillance, tracking and misuse of personal information. To address this, the order supports explicitly emerging techniques like differential privacy, federated learning and on-device processing that allow training AI models without aggregating raw user data in centralized repositories.
It directs NIST to issue guidance identifying suitable privacy-enhancing techniques for different AI use cases based on data sensitivity. Federal agencies must evaluate their AI data practices and strengthen privacy protections within fixed timelines. New programs will fund research into privacy-preserving technologies.
Assessing and Mitigating Algorithmic Biases
Given AI’s risk of perpetuating discrimination through opaque algorithms, the order requires evaluating specific biases and unfair impacts in automated systems used for criminal justice, lending, employment, education admissions and other high-stakes decisions that shape people’s lives.
Both existing and new systems must be rigorously tested on diverse datasets to uncover biases against protected groups across race, gender, age, disability status and other attributes. The results will inform policy changes to counter discrimination and expand opportunity.
Supporting AI Innovation Alongside Democratic Values
To accelerate AI discoveries, the order aims to streamline visa processes to attract global talent and fund expanded access to computing resources, datasets and educational programs for researchers. It instructs relevant agencies to devise plans to achieve this within fixed timeframes.
Moreover, the order emphasizes that any breakthroughs must be ethical and aligned with our values. For example, it prohibits high-risk systems, such as certain types of social scoring, without adequate due process safeguards. Prioritizing ethical innovation is essential. Additionally, it is critical to balance oversight and flexibility to allow for startup innovation. While big tech welcomes talent incentives, smaller developers are concerned about the burdens. Academia also wants more research funding. The order addresses a variety of interests, but the key is to put principles into practice.
Preparing for AI Impacts on Jobs and Skills
The order takes a proactive approach to address AI’s potential displacement of jobs across industries. It directs the Secretary of Labor to lead an interagency working group to study these impacts on the workforce.
Specifically, it mandates examining displacement effects across occupations, sectors, income levels, age groups, education levels, and locations. The findings will identify vulnerable segments of workers needing assistance.
Based on the insights, within 180 days, the working group must formulate an action plan detailing provisions to assist affected workers. This includes facilities like transition support programs, job search aid, skills training grants and modernizing the social safety net to support career shifts.
Robust implementation of these directives will ensure that vulnerable workers are protected. The order also tasks agencies to jointly tackle AI’s potential impacts on civil rights and national security. Overall, it puts in place mechanisms for mitigating risks to jobs, fairness and safety.
But realization depends on coordination between Labor, Education, Commerce, Defense and more. With proper funding and follow-through, the threats could be addressed. Done right, the order provides a roadmap to smoothly transition the workforce for the future.
BLETCHLEY, ENGLAND – NOVEMBER 2: Prime Minister Rishi Sunak (right) meets with President of the … [+] European Commission Ursula von der Leyen during the AI safety summit, the first global summit on the safe use of artificial intelligence, at Bletchley Park on November 2, 2023 in Bletchley, England. The UK Government are hosting the AI Safety Summit bringing together international governments, leading AI companies, civil society groups and experts in research to consider the risks of AI, especially at the frontier of development, and discuss how they can be mitigated through internationally coordinated action. (Photo by Joe Giddens – WPA Pool/Getty Images)
Getty ImagesFostering Global Cooperation on AI
AI’s risks and opportunities require worldwide cooperation, as unilateral approaches could undermine innovation. To enable allies to reap the gains of AI collectively, the order instructs the State Department to organize a Global AI Partnership for Democratic Advancement within 90 days.
This partnership will convene foreign ministers, technology regulators, philanthropies, academics, civil society groups and companies to align strategies on AI governance across nations based on shared values. The goal is to foster responsible development and prevent a race to the bottom.
It will collaborate on creating new international technical standards, risk assessment methodologies, safety labels and incentives for trustworthy AI via multi-stakeholder organizations like the IEEE and ISO. Public-private partnerships will provide low-cost expertise to nations lacking capabilities.
The order also proposes establishing a new multilateral AI research institute to illuminate high-uncertainty issues like the societal impacts of generative AI through global knowledge sharing. Avoiding unilateral competition in AI via cooperation is wise.
Lastly, the partnership model for developing standards provides a lightweight, flexible framework compared to legislation. This collaborative approach navigates the complex political terrain and varying interests across different factions including industry, academia, government and the public. The order’s expansive scope covers a wide range of concerns in an effort to set a broad foundation for cooperation.
Implementation Challenges and Stakeholder Reactions
There are concerns about practical implementation given the extensive workload imposed on agencies. Past executive orders have faced challenges in execution. Impacted groups like startups feel sidelined by burdens while academics want more funding.
However, easing visa rules to attract foreign AI talent is a positive step welcomed by tech companies. But issues like job losses, civil rights impacts, and national security require continued vigilance across agencies. Realizing impact depends on turning principles into on-ground practices.
The Way Forward
This expansive Executive Order is a marker signaling that America aims to lead the world in developing AI that respects democratic values while pragmatically mitigating risks.
However, realizing this vision requires multi-sector coordination and tangible action. Some vital next steps include:
Formulating ethics codes and risk assessment tools for trustworthy AI tailored to specific industries.
Investing in research and commercialization of privacy-enhancing and bias-mitigating technologies.
Expanding education in not just technical but also ethical aspects of AI.
Developing mechanisms for ongoing community input into AI policies and grievance redressal.
Building global consensus on AI safety standards through strategic diplomacy.
Tracking critical metrics on AI fairness, safety, and accessibility to keep systems accountable.
Effective execution will determine the order’s real-world impact. As Ceze points out, “Developing open standards for model evaluation is a great idea. But benchmarking models are a moving target. Models are iterative and evolve quickly as they ingest more and more data. If we are to succeed in evaluating models, it will have to be done continuously—which is not a common practice at the moment.”
With careful implementation, this order could pioneer precedent-setting guardrails against AI risks that earned public trust and shaped development trajectories worldwide. Getting there needs all hands on deck.
{Categories} *ALL*,_Category: Implications{/Categories}
{URL}https://www.forbes.com/sites/markminevich/2023/11/02/bidens-bold-ai-blueprint-navigating-ethics-and-innovation/{/URL}
{Author}Mark Minevich, Contributor{/Author}
{Image}https://imageio.forbes.com/specials-images/imageserve/6543942465cc26a199ac16d6/0x0.jpg?format=jpg&height=600&width=1200&fit=bounds{/Image}
{Keywords}AI,/ai,Innovation,/innovation,AI,/ai,standard{/Keywords}
{Source}Implications{/Source}
{Thumb}{/Thumb}