European Commission applies human intelligence to regulate artificial intelligence –
what are the implications for automated vehicles?
by Nick Reed
Following an extensive stakeholder consultation process and the involvement of a ‘High- Level Expert Group’, the European Commission recently published its proposed approach to the regulation of artificial intelligence (AI).
It recognises the potential for AI to bring a wide array of economic and societal benefits but that the attributes that deliver socio-economic benefits can also create new risks or negative consequences for individuals or society. The proposal therefore sets out a methodology defining what it describes as “high-risk” AI systems – those that pose significant risks to health and safety and/or the fundamental rights of persons – and recommends a regulatory approach involving EU-wide legislation for higher risk AI systems and codes of conduct for non-high-risk systems. This would place requirements on developers of high-risk AI systems to ensure high-quality data, documentation, testing, traceability, transparency, human oversight, accuracy, risk management and robustness.
In particular, it specifically describes requirements for data logging to ensure AI systems perform appropriately according to their intended purpose, that their operation is sufficiently transparent to enable users to interpret the system’s output (and use it appropriately) and that system operators can enable a competent national authority to access the data in response to reasonable request.
The proposed regulation also lists penalties to be imposed for organisations found guilty of performing practices prohibited by the proposed regulation. These can be up to (the greater of) €30m or 6% of global turnover, depending on the severity of the transgression.
Although the proposal does not discuss automated vehicles directly, their classification as a high-risk AI system seems to be captured in Annex III to the proposal, which under the category “Management and operation of critical infrastructure”, lists AI systems used as safety components for the “operation of road traffic”.
The European Commission’s proposed approach to AI complements twenty recommendations published by the European Commission from an Expert Group (of which I was one) on the ethics of connected and automated vehicles. This is a topic that occasionally suffers from a misplaced emphasis on thought experiments around how AVs should behave in dilemma situations related to the original ‘trolley problem’. However, the expert group explored broader ethical concerns associated with the development and deployment of AVs, focused on three categories; safety, transparency and liability. The group’s recommendations make similar points to the broader AI proposal, citing a need for data to provide assurance around safety and noting that privacy concerns must be addressed in the collection and storage of any sensitive data. The recommendations also consider the circumstances in which an AV might be expected to break existing traffic rules in the interests of safety, highlighting the need for evidence to explain why the AV acted in such a way.
AVs have the potential to deliver huge safety, economic and accessibility benefits but there is no escaping that, to have confidence they operate safely in public environments, we will need verifiable data to show that they are suitably perceiving the world around them, making correct decisions about how to operate in response, competently enacting those decisions through inputs to vehicle actuators and accurately recording outcomes that demonstrate safe compliant and considerate driving. The European Commission’s proposed regulations for AI and the expert group’s recommendations on CAV ethics highlight that if manufacturers and developers seek to capitalise on the opportunities that AVs offer, then the topic of safe, secure, standardised data collection must be addressed. Other industries, such as the aviation sector, have successfully overcome this challenge. We now need the automotive and technology industries and public sector regulators to work collaboratively to agree suitable protocols that unlock the anticipated benefits of AVs whilst also protecting the fundamental rights of individuals and intellectual property of developer organisations. This application of human intelligence will be the gateway to extracting maximum value from its artificial counterpart.
Is your AI business struggling to navigate the current atmosphere?
Then get in touch with us for a free consultation.
Email us at firstname.lastname@example.org with any questions or queries
Don’t hesitate to contact us and subscribe to our Newsletter.