Why Hawaiʻi Needs To Take An Aggressive Approach To Regulate AI
We are heading soon to the point where artificial intelligence can do anything that humans can do, but faster.
November 26, 2023 · 6 min read
About the Authors
Tam Hunt is a lawyer and activist based on the Big Island. He is co-founder of Think B.I.G. and a board member for the Hawaii Electric Vehicle association.
Noel Morin is a climate, sustainability, and resilience advocate based in Hawaii.
Argentina just had the world’s first “AI election.” No, AI was not running for office, but politicians used artificial intelligence very aggressively to create deepfake videos and images that seemed to have an impact on the presidential election.
The New York Times describes how AI has made candidates say things they did not, and put them in famous movies and memes. It has created campaign posters and triggered debates over whether real videos are actually real.
A lot is happening as the world reels from the dramatic improvements in AI in the last year. AI “chatbots,” image and video generators, recommendation engines, and many other forms of AI are getting scary good, and improving scarily fast. Human-level jobs are starting to be done by AI and this trend will accelerate in the next few years.
We are heading rather soon for the point where AI can do anything that humans can do — but far, far faster. This level of AI is known as “Artificial General Intelligence” or AGI.
Industry predictions for when AGI will arrive have advanced from “maybe never” to “sometime this century” to now as early as 2028, just five years away.
Artificial intelligence is scarily real and so must be regulated by government. (Wikimedia Commons/Alenoach/2023)
An academic paper from earlier this year concluded after extensive testing of OpenAI’s GPT4.0 chatbot that it showed “sparks of AGI” already. Their study made big news after they found that this AI could already do better on the Uniform Bar Exam (the standardized portion of the bar exam to become a certified lawyer) than 90% of human test takers — up from just 10% in the previous version.
World leaders met in November in England to discuss how to keep AI safe. They produced a document, the Bletchley Declaration, that was signed by all nations present, including China. This is a good start, but we will, of course, need far more to ensure that AI remains safe and to avoid an out-of-control AI arms race between nations.
At the national level, President Joe Biden also issued a (very long) executive order this month that provides a framework to keep AI safe in the U.S. It requires changes to several federal agencies and puts in place various safeguards.
This is also a good start, and we applaud Biden for acting quickly. It is, however, only an executive order, and the next president could erase it in a moment. We need legislation and a new federal agency to robustly regulate AI.
Senate Majority Leader Chuck Schumer is working on a broader piece of legislation, but it is not at all clear whether such a bill will make it through Congress during this very strange election cycle. It is also unlikely that Congress will pass anything that is aggressive enough or timely enough, given the very rapid pace of AI development.
States can act more quickly and more aggressively to meet this urgent need for regulating AI. Hawaii is leading the way with a new draft bill co-sponsored by Sens. Mike Gabbard and Chris Lee.
This bill is groundbreaking in that it places the “precautionary principle” front and center in allowing AI products to be used here in Hawaii. The bill would, under this principle, require proactively that AI companies demonstrate the safety of their products before they’re allowed to be used in Hawaii.
The bill also adopts the European Union’s approach of “risk-based regulation” in that the height of the bar to be cleared by each company will be based on the potential dangers of each technology. The greater the risk, the higher the bar that must be cleared.
States can act more quickly and more aggressively to meet this urgent need.
For AGI products — let’s say OpenAI GPT-5 or 6, for example, when they become available — the bar will be very high indeed for demonstrating safety. But for less risky products such as product recommendation algorithms (like Netflix or YouTube uses), the bar will be lower.
There is a good precedent for what this bill is trying to do. Hawaii has strict cryptocurrency regulations already in place. While Hawaii can’t outlaw the use of cryptocurrency entirely, it can regulate which crypto exchanges are allowed to operate here and which exchanges Hawaii residents can access.
Hawaii adopted strict rules in 2016 to protect consumers from unscrupulous crypto exchanges — risks that have been highlighted by the recent conviction of Sam Bankman-Fried for numerous financial crimes based on his time running FTX, once a major crypto exchange before it collapsed.
Under Hawaii’s current rules, only 15 crypto exchanges are allowed to operate in Hawaii.
Hawaii should take a similar approach to AI websites like OpenAI’s ChatGPT, Bing (which uses a version of ChatGPT), Anthropic, Google’s Bard, etc. The Gabbard and Lee draft bill would achieve this.
Agents Of AI
We are already trending toward AI “agents” as the next generation of AI tools. While today’s AI tools like ChatGPT can do many impressive things in terms of mastering human language, producing realistic images, and even coding complex software, they are still not autonomous agents. They require human input. The next generation of AI tools, however, will be autonomous and capable of executing complex and lengthy tasks without human input.
Since computing power is cheap and easily accessible, these AI agents will soon flood the Internet in the billions and trillions, wreaking havoc as they go — unless they are very tightly regulated. The Cambridge Analytica scandal from 2016, which used early AI tools to individually target social media users with election misinformation, will look quaint in comparison to what is coming soon in terms of AI agents.
Argentina’s example, as the first “AI election,” is just the very beginning of what we can expect to see.
Hawaii can and should take the lead on this highly important issue by ensuring that new AI tools demonstrate their safety before they’re released in Hawaii.
Please contact your legislators to share your thoughts on AI and AI regulation.
Community Voices aims to encourage broad discussion on many topics of community interest. It’s kind of a cross between Letters to the Editor and op-eds. This is your space to talk about important issues or interesting people who are making a difference in our world. Column lengths should be no more than 800 words and we need a photo of the author and a bio. We welcome video commentary and other multimedia formats. Send to news@civilbeat.org. The opinions and information expressed in Community Voices are solely those of the authors and not Civil Beat.
