Open AI CEO Sam Altman’s testimony before the Senate Judiciary Subcommittee on Privacy, Technology, and Law yesterday illustrates the importance of generative artificial intelligence (AI) and the subtleties surrounding its development. ChatGPT, like tools launched by other companies, democratized technology by bringing tremendous computing power to search, data analysis, video and audio generation, software development, and many other areas. . Generative AI has the power to change the way people find information, generate new audio and video, create new products, and respond to new events in real time.
At the same time, however, it also raises several issues for consumers, academics and policy makers. Worrying issues include harmful content, disinformation, political favoritism, racial bias, lack of transparency, workforce impact, and intellectual property theft. Altman’s testimony, along with testimony from IBM vice president Christina Montgomery and New York University professor Gary Marcus, provides an opportunity to explain generative AI and raise lawmakers’ concerns about its social, economic and electoral implications. given the opportunity to express
This post reviews key takeaways from the Senate hearings and next steps for emerging technologies. In general, there was broad agreement on the risks of disinformation, biased decision-making, invasion of privacy, and job losses. Most of the speakers believed that these issues were very serious and needed meaningful action. Even more surprising was the frequent rise of bipartisan calls for tighter regulation and greater disclosure of AI use. At a time when much of the conversation in Washington has become quite polarized and partisan, most lawmakers are convinced that stronger guardrails are needed to protect fundamental human values, especially in the areas of: Agreed.
harmful effects
As AI becomes more pervasive in many fields, there is widespread agreement about its harmful content and worrying effects. Generative AI brings sophisticated algorithms to the public in the form of online prompts and templates, allowing virtually anyone to create and spread false narratives. Many expect the 2024 election to be a tsunami of misinformation, but this was because the election was so close that many people and organizations were forced to misrepresent themselves with little regard for fairness or factual accuracy. , which motivated them to create videos, fake audio, and accusatory writings.
Several senators spoke of issues of stigma, loss of privacy and potential job losses for those whose tasks could be replaced by AI algorithms. There is currently a lack of transparency about how the AI works, the training data it is based on, and how the AI makes decisions. Many were concerned about how AI would fuel mass manipulation and make people think and act in distorted ways.
Subcommittee chair Richard Blumenthal (D., Connecticut) worried about “sound duplicating software” and a dystopia “no longer a sci-fi fantasy.” He cited the weaponization of disinformation, housing discrimination and deepfake videos as his top concerns. Given these concerns, ranking member Josh Hawley (R., Missouri) complained about the “power of the few” and the need to balance innovation and moral responsibility.
Stricter disclosure
Less expected was the fact that nearly all legislators and outside speakers, including industry leaders, called for stricter disclosure requirements. Altman and other speakers said consumers need to be alerted when generative AI is used to create video, audio, and other types of products. People need to know what is human-generated and what comes from algorithms, as it can affect how they view certain products.
independent audit
Since AI is still in the early stages of testing, it’s important to monitor model results externally. Large language models should be tested through a third party and the results made publicly available. This helps people understand how AI is performing and which applications are particularly problematic. Marcus of New York University noted the usefulness of nutrition labels and said such labels could benefit AI products and services.
RISK-RELATED LIMITATIONS
IBM’s Montgomery argued for risk-related regulation, noting that regulation may differ according to risk. His high-risk AI required more scrutiny than low-risk applications. Incidents involving human safety, biomedical risk, or harm to specific populations require more detailed analysis, monitoring, inspection, and public review. These applications require more detailed investigation as they can have very disastrous consequences.
License requirements
Some human activities are considered dangerous enough by society to require a license, such as driving a car, hunting, fishing, or starting a business. Some speakers asked whether AI licensing would help mitigate potential harm and increase accountability among technology providers. Several senators also felt that, especially for high-risk AI applications, licensing requirements would be beneficial as long as these rules didn’t restrict open source models and small businesses.
Ethics Review Board
IBM’s Montgomery noted that the company has hired AI ethics experts, overseen the company’s product development, and established an AI review board to evaluate its response to key human principles. Implementing such mechanisms will ensure that the human factor is put at the forefront and that AI products receive meaningful internal oversight.
Intellectual property protection
Senator Marsha Blackburn stressed the importance of intellectual property protection, saying creators should be compensated when their music, voice and content are used to train AI. She gave an example of using AI to create songs like Garth Brooks, resulting in what sounds like his song “Simple Man.”
new regulator
Faced with these various proposals was the theme of whether the United States needs a new AI regulator. The proliferation of automobiles prompted the creation of the National Highway Safety Administration, and the advent of television and radio gave birth to the Federal Communications Commission.
Develop a new agency with the technical expertise to monitor, evaluate and regulate AI, as AI innovation spans so many different sectors and has so much impact on governments, businesses and consumers. Maybe it’s time. Failing that, sector-specific solutions are needed, with different AI rules for finance, health, transportation, education, housing, and employment. Its “Tower of Babel” approach is disorganized, unorganized and ineffective in combating the ill effects of AI, and many will be left unsatisfied with the results.
