Pages

9 January 2024

Making sense of 2023’s tsunami of AI policy changes

ROGER COCHETTI

Few events better characterized 2023’s tsunami of pronouncements about regulating artificial intelligence (AI) than Pope Francis’s plea for the world. 2023 saw thousands of AI policy pronouncements, proposals, laws, orders and regulations, as well as an avalanche of headlines, talking heads, hearings, conferences, editorials and scholarly publications. It’s no wonder that Pew Research reports a majority of Americans are worried about AI.

Underlying this are numerous fears — some generated by a century of science fiction stories about robots running amok and others generated by documented problems such as students using AI to cheat, mass AI copyright infringement, realistic AI-generated impersonations of officials, AI-controlled cars getting into accidents, people losing jobs as they are replaced by AI, and more. Media hyperbole has left many wondering whether humanity is at the dawn of a new “AI age” comparable to the Bronz Age or the Industrial Revolution.

A little background is probably useful. (Full disclosure: I was on the executive team of an early-mid 1990s IBM experiment to launch a consumer-facing AI business. The technology wasn’t ready, so the business was quietly folded, although within a year, IBM’s Deep Blue AI beat world chess champion Gary Kasparov in a famous chess match.)

First, not only is there no agreed-upon definition of AI, there is no agreement as to whether there is an agreed definition.

When asked whether there is an official definition of AI, Google’s Bard responds, “There isn’t one single, universally accepted definition. … However, there are some general concepts that most definitions of AI agree on.” Conversely, Microsoft-affiliated ChatGPT responds “Yes, AI refers to the development of computer systems capable of performing tasks that typically require human intelligence, such as….” Although the 38-member-nation Organization for Economic Cooperation and Development has worked for years to define artificial intelligence, AI experts from these countries (which are mostly European and do not include China, Russia or India) have offered an evolving definition that is accepted as “universal by some countries.

Second, regardless of one’s definition, some form of AI has been around for a while. As far back as the 1950s, computers were used for large data trend analyses, and by the 1960s, MIT was testing a generative computer service that could act like a human and exchange messages. Closed computer services that could generate simple text messages (what we today call “chatbots”) gradually grew through the 1990s when they merged with the Internet, making access to computer chatting universal.

Importantly, it took vast increases in processing and the unlimited ability of the Internet to allow computers to instantly scan millions of databases, websites and other resources managed by complex new software to create the large language generative AI that captivated policy-makers in 2023. In response, hundreds of provincial and local governments enacted AI laws or regulations; the United Nations Security Council held a special meeting on AI; Britain convened a global summit conference on AI where hundreds of inter-governmental meetings and conferences on “AI policy” took place; and over a dozen AI policy scholarly centers, a half-dozen major industry coalitions and NGO/labor union projects were created.

Third, the White House has issued a series of proclamations, orders and directives to federal agencies instigating an enormous effort to develop American AI guidelines and regulations. Similarly, China has enacted a major AI law enabling its Cyberspace Commission and other agencies to regulate AI, and the European Union has provisionally approved a comprehensive “AI Act” setting in motion a broad regulatory program among both EU organs and its 27 national governments. Separately, the issue of legal liabilities for alleged AI harms has begun to appear before law courts everywhere, and it will inevitably grow. Even Supreme Court Chief Justice John Roberts highlighted legal uncertainties caused by AI in his recent annual report.

Finally, 2023 saw the first major successful labor union actions demanding AI terms and conditions in union contracts — but by no means the last. The use of AI will impact many employees and their unions will demand AI terms be included in union contracts. These will likely parallel AI terms that will be incorporated into many intellectual property and similar licensing agreements.

A global AI policy juggernaut was launched in 2023 but there is no precedent for its breadth, depth and speed. Thousands of regulatory and judicial proceedings are emerging in hundreds of jurisdictions, making uneven AI regulation almost certain. Although the U.S. proposed setting global AI policy principals at the United Nations, given the competitive nature of the industry and the increasingly contentious relationships on this topic among the leading geopolitical players — the U.S., China and the EU — it’s difficult to see harmonization anytime soon.

No comments:

Post a Comment