This post features the second episode of the JohnWright.ai Artificial Intelligence Podcast. Episode 2 is titled Building Behaviorally Smarter Cities with Spatial featuring Lyden Foust as my guest. Lyden is the co-founder of Spatial, a company that utilizes artificial intelligence technology combined with principles of ethnography to collect and analyze public expressions on social networks in order to better understand how humans interact with their surroundings in cities. Spatial examines the human layer – an invisible layer of data that can be derived from conversations on social networks that contain people’s insights and opinions about the cities that they live in.
On December 12, 2017, members of Congress – in a bipartisan effort – introduced bills to enact the Fundamentally Understanding The Usability and Realistic Evolution (FUTURE) of Artificial Intelligence Act of 2017. In the House of Representatives, Rep. John Delaney (D-MD) – founder of the Artificial Intelligence (AI) Caucus – along with six cosponsors – including AI Caucus co-chair Pete Olson (R-TX) – introduced H.R.4625. In the Senate, Sen. Maria Cantwell (D-WA) along with co-sponsors Sens. Todd Young (R-IN) and Edward Markey (D-MA) introduced S.2217.
In a press release about the bills, Sen. Markey is quoted as stating,
“While artificial intelligence holds the promise of providing goods and services more efficiently and effectively, increased automation has potentially broad negative impacts on our workforce and our privacy[.] This bill serves as an important step in bringing together all stakeholders to better understand how this new technology will impact our lives.”
Under the bills, stakeholders would be brought together by requiring the Department of Commerce to establish a 19-member Federal Advisory Committee (FAC) on the Development and Implementation of Artificial Intelligence. The FAC would be comprised of a geographically diverse group of members from various backgrounds including academia and private industry, along with members from labor organizations and civil society – including members from groups advocating for civil rights.
The FAC would be expected to study issues related to artificial intelligence and would also provide relevant advice including a formal report to the Secretary of Commerce and Congress with recommendations on potential administrative or legislative action relating to artificial intelligence. Included in the FAC’s priorities would be the development of guidance or recommendations on four policy objectives identified in the bills, being:
(A) to promote a climate of investment and innovation to ensure the global competitiveness of the United States;
(B) to optimize the development of artificial intelligence to address the potential growth, restructuring, or other changes in the United States workforce that results from the development of artificial intelligence;
(C) to promote and support the unbiased development and application of artificial intelligence; and
(D) to protect the privacy rights of individuals.
Notably, the bills attempt to create a legal definition of artificial intelligence. This is both interesting and important because there is no widely-accepted legal definition of artificial intelligence, and there needs to be one before there is any effort to institute regulation. The bills attempt to define artificial intelligence as:
(A) Any artificial systems that perform tasks under varying and unpredictable circumstances, without significant human oversight, or that can learn from their experience and improve their performance. Such systems may be developed in computer software, physical hardware, or other contexts not yet contemplated. They may solve tasks requiring human-like perception, cognition, planning, learning, communication, or physical action. In general, the more human-like the system within the context of its tasks, the more it can be said to use artificial intelligence.
(B) Systems that think like humans, such as cognitive architectures and neural networks.
(C) Systems that act like humans, such as systems that can pass the Turing test or other comparable test via natural language processing, knowledge representation, automated reasoning, and learning.
(D) A set of techniques, including machine learning, that seek to approximate some cognitive task.
(E) Systems that act rationally, such as intelligent software agents and embodied robots that achieve goals via perception, planning, reasoning, learning, communicating, decision making, and acting.
The bills also delineate between artificial general intelligence versus narrow artificial intelligence. The bills define artificial general intelligence as a “notional future” system, whereas narrow artificial intelligence is defined as a presently-existing system. This suggests that Congress is well-informed as to the state of artificial intelligence today and is deciding to take a pragmatic approach. It is no surprise then to learn that the bills were introduced the same day that the Senate Committee on Commerce, Science and Transportation subcommittee on Communications, Technology, Innovation and the Internet held a hearing with a panel of witnesses providing testimony on the current state of artificial intelligence technology.
The bills have both been referred to various committees in both the House and the Senate, so there promises to be further discussion about a possible federal advisory committee focused on artificial intelligence. Do you think that the United States needs such an advisory committee? Should the government do more or less when it comes to the development of artificial intelligence technologies? Please share your thoughts in the comments section below.
This post features the first episode of the JohnWright.ai Artificial Intelligence Podcast. Episode 1 is titled “Autonomous Vehicles and Robot Cars” featuring Damian Porcari as my guest.
Damian recently retired as Director of Licensing and Enforcement for Ford Global Technologies LLC, a wholly-owned subsidiary of Ford Motor Company. Damian spent 28 years at Ford as part of the in-house legal department and held other positions such as Director of Technology Commercialization and Intellectual Property Managing Counsel.
Damian began his career at Ford in the position of Senior Patent Counsel after spending seven years in the U.S. Army legal department holding such positions as Intellectual Property Attorney and Senior Attorney for Software. Damian is currently the owner and founder of International Design Law PC, a law firm that specializes in protecting designs through registrations and design patents.
Volvo Cars announced in a November 20, 2017 press release that it signed a non-exclusive framework agreement with Uber (in the headlines for other reasons as of this writing) to supply the technology company with tens of thousands of autonomous XC90 sport utility vehicles. This announcement builds on the alliance between Uber and Volvo Cars announced last year. While an actual, binding purchase agreement may still be in the works, what this headline suggests is that technology companies like Uber are aggressively preparing to launch fleets of autonomous vehicles onto American streets as soon as 2019.
Where, exactly, such a large scale launch could occur in the United States as of today is largely dependent on a patchwork of state laws and municipal ordinances that have been implemented throughout the United States (organized neatly by Gabriel Weiner and Bryant Walker Smith here: Automated Driving: Legislative and Regulatory Action). There is, however, legislation pending in Congress pertaining to autonomous vehicles that could preempt some state and local laws and provide a cohesive framework for the development of autonomous vehicle technology.
On September 6, 2017, the U.S. House of Representatives passed the Safely Ensuring Lives Future Deployment and Research In Vehicle Evolution Act (SELF DRIVE Act) H.R. 3388. Introduced by Rep. Robert Latta (R-OH) and initially co-sponsored by Rep. Janice Schakowsky (D-IL) in July 2017 (picking up 30 additional co-sponsors the day before it passed), the SELF DRIVE Act has been a bi-partisan effort that sailed through the House with relative speed, a legislative anomaly in recent history.
The U.S. Senate is currently considering a separate bill called the American Vision for Safer Transportation through Advancement of Revolutionary Technologies Act (AV START Act) S.1885. Introduced on September 28, 2017, by Sen. John Thune (R-SD) and co-sponsored by Sen. Gary Peters (D-MI), Sen. Roy Blunt (R-MO) and Sen. Debbie Stabenow (D-MI), the AV START Act also represents a bipartisan effort.
Both bills share some similarities. Notably, neither applies to commercial vehicles. So, for example, a federal statute resulting from either bill would not apply to Tesla’s recently-announced Semi truck. Also, both bills reference SAE International’s J3016 Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles. The taxonomy sets up five different levels of autonomy in vehicles from Level 0 for no autonomy to Level 5 being completely autonomous. Even though Level 5 autonomous vehicles are not on the road today, this achievement is anticipated soon and neither bill absolutely requires a human backup driver in an autonomous vehicle. Finally, both bills include identical federal preemption language allowing states and municipalities to enact and enforce laws pertaining to autonomous vehicles only so long as they are consistent with federal laws and regulations.
Some federal regulations currently in place are written in such a way that the only way to comply is by having a human driver behind the wheel of a vehicle. Since some autonomous vehicles will have no driver at all, these regulations could stand in the way of deploying such vehicles. Both bills thus require a review of the Federal Motor Vehicle Safety Standards (FMVSS) to resolve these issues. The House bill requires a review of the FMVSS to begin within 180 days of enactment, whereas the Senate bill takes it a step further by requiring a completed report on conflicting provisions within 180 days of enactment, and commencement of a U.S. Department of Transportation (USDOT) rulemaking 90 days later.
There are other provisions in the two pieces of proposed legislation where the Senate version – perhaps having the benefit of hindsight – takes a slightly different approach on the same or similar issue. For example, both bills address cybersecurity, but the Senate version is more broad. Where the House bill merely requires manufacturers to have a cybersecurity plan with a written policy and appointed officer, the Senate bill takes the same concept further by adding additional requirements to the plan, allowing for federal and public inspection of such plans and calling for a method for security researchers to confidentially share information about vulnerabilities in autonomous vehicle systems.
Rather than allowing for an unlimited number of autonomous vehicles to hit the road once the technology is available, both bills create step increases of the number of autonomous vehicles that can be sold in successive twelve-month periods by creating exemptions. Where the House bill allows for 25,000; 75,000, and; 100,000 exempted autonomous vehicles during the first three years, respectively, the Senate bill starts at 50,000 during the first year and allows for the possibility of exemptions exceeding 100,000.
And, of course, both the House and the Senate bills focus heavily on safety regulations, with the House version requiring a USDOT rulemaking on safety assessment certification, and the Senate version requiring safety evaluation reports by manufacturers. These safety provisions are interesting, but they are not unexpected. The federal government typically regulates safety of actual vehicles whereas state governments regulate how vehicles operate on roads within state lines. More interesting is how the bills address – or refrain from addressing – important privacy and ethics concerns.
There is a notable difference in how the two bills address privacy. The House version requires an autonomous vehicle manufacturer to establish a written privacy plan. In contrast, the Senate omits that requirement. In recent guidance issued after the passage of the SELF DRIVE Act and before the introduction of the AV START Act, the National Highway and Traffic Safety Administration (NHTSA) acknowledges that privacy and ethical matters are important to deliberate. (See: Footnote 1). However, pointing to its website, NHTSA takes the approach that privacy matters are under the purview of the Federal Trade Commission (FTC). (See: What is NHTSA’s approach to privacy?). As to ethics, it is NHTSA’s position that “there is no consensus around acceptable ethical decision-making given the depth of the element is not yet understood nor are there metrics to evaluate against.” (See: What is NHTSA’s approach to ethical considerations?).
There is certainly more interesting deliberation to come regarding privacy and ethics. Both bills set up advisory committees consisting of experts in various fields. The House version establishes a new autonomous vehicle Advisory Council, where the Senate establishes a new Technical Committee. While the titles of these groups are different, both will be expected to address pressing issues relating to privacy and ethics in addition to technical and safety concerns.
This blog will follow these discussions and more, so please provide your comments and check back later for further updates.