Policymakers want to regulate AI but lack consensus on how

3 years ago 261

Commentary: AI is considered "world changing" by policymakers, but it's unclear however to guarantee affirmative outcomes.

According to a caller Clifford Chance survey of 1,000 tech argumentation experts crossed the United States, U.K., Germany and France, policymakers are acrophobic astir the interaction of artificial intelligence, but possibly not astir enough. Though policymakers rightly interest astir cybersecurity, it's possibly excessively casual to absorption connected near-term, evident threats portion the longer-term, not-obvious-at-all threats of AI get ignored.

Or, rather, not ignored, but determination is nary statement connected however to tackle emerging issues with AI.

SEE: Artificial quality morals policy (TechRepublic Premium)

AI problems

When YouGov polled tech argumentation experts connected behalf of Clifford Chance and asked precedence areas for regularisation ("To what grade bash you deliberation the pursuing issues should beryllium priorities for caller authorities oregon regulation?"), ethical usage of AI and algorithmic bias ranked good down the pecking bid from different issues:

94%—Cybersecurity
92%—Data privacy, information extortion and information sharing
90%—Sexual maltreatment and exploitation of minors
86%—Misinformation / disinformation
81%—Tax contribution
78%—Ethical usage of artificial intelligence
78%—Creating a harmless abstraction for children
76%—Freedom of code online
75%—Fair contention among exertion companies
71%—Algorithmic bias and transparency
70%—Content moderation
70%—Treatment of minorities and disadvantaged
65%—Emotional wellbeing
65%—Emotional and intelligence wellbeing of users
62%—Treatment of gig system workers
53%—Self-harm

Just 23% complaint algorithmic bias, and 33% complaint the ethical usage of AI, arsenic a apical precedence for regulation. Maybe this isn't a large deal, but that AI (or, much accurately, machine learning) finds its mode into higher-ranked priorities similar information privateness and misinformation. Indeed, it's arguably the superior catalyst for problems successful these areas, not to notation the "brains" down blase cybersecurity threats.

Also, arsenic the study authors summarize, "While artificial quality is perceived to beryllium a apt nett bully for nine and the economy, determination is simply a interest that it volition entrench existing inequalities, benefitting bigger businesses (78% affirmative effect from AI) much than the young (42% affirmative effective) oregon those from number groups (23% affirmative effect). This is the insidious broadside of AI/ML, and thing I've highlighted before. As elaborate successful Anaconda's State of Data Science 2021 report, the biggest interest information scientists person with AI contiguous is the possibility, adjacent likelihood, of bias successful the algorithms. Such interest is well-founded, but casual to ignore. After all, it's hard to look distant from the billions of idiosyncratic records that person been breached.

But a small AI/ML bias that softly guarantees that a definite people of exertion won't get the job? That's casual to miss.

SEE: Open root powers AI, yet policymakers haven't seemed to notice (TechRepublic)

But, arguably, a overmuch bigger deal, due to the fact that what, exactly, volition policymakers bash done regularisation to amended cybersecurity? Last I checked, hackers interruption each sorts of laws to ace into firm databases. Will different instrumentality alteration that? Or however astir information privacy? Are we going to get different GDPR bonanza of "click present to judge cookies truthful you tin really bash what you were hoping to bash connected this site" non-choices? Such regulations don't look to beryllium helping anyone. (And, yes, I cognize that European regulators aren't truly to blame: It's the data-hungry websites that stink.)

Speaking of GDPR, don't beryllium amazed that, according to the survey, policymakers similar the thought of enhanced operational requirements astir AI similar the mandatory notification of users each clip they interact with an AI strategy (82% support). If that sounds a spot similar GDPR, it is. And if the mode we're going to woody with imaginable problems with the ethical usage of AI/bias is done much confusing consent pop-ups, we request to see alternatives. Now.

Eighty-three percent of survey respondents see AI "world changing," but nary 1 seems to cognize rather however to marque it safe. As the study concludes, "The regulatory scenery for AI volition apt look gradually, with a substance of AI-specific and non-AI-specific binding rules, non-binding codes of practice, and sets of regulatory guidance. As much pieces are added to the puzzle, determination is simply a hazard of some geographical fragmentation and runaway regulatory hyperinflation, with aggregate akin oregon overlapping sets of rules being generated by antithetic bodies."

Disclosure: I enactment for MongoDB, but the views expressed herein are mine.

Data, Analytics and AI Newsletter

Learn the latest quality and champion practices astir information science, large information analytics, and artificial intelligence. Delivered Mondays