Washington
CNN
—
The AI policy think tank has warned the U.S. government against OpenAI and its massive popularity, arguing that the technology’s propensity to produce algorithmic biases, privacy concerns and occasional inaccurate results could violate federal consumer protection laws. of GPT artificial intelligence products seeking investigation.
The US Federal Trade Commission should ban OpenAI from releasing future versions of GPT and establish new regulations for the burgeoning AI sector, the Center for AI and Digital Policy (CAIDP) said Thursday in an agency complaint. rice field.
The complaint seeks to fully counter the FTC’s broad consumer protection powers against what CAIDP has portrayed as a runaway western experiment in which consumers pay for the unintended consequences of AI development. there is And as tech-skeptical officials, such as FTC Chairman Rina Khan, warn of the dangers of uncontrolled data use for commercial purposes and new ways by tech companies, this is a direct threat to AI. It could be an early test of the U.S. government’s willingness to regulate. Trying to establish a monopoly.
The FTC declined to comment. OpenAI did not respond to a request for comment.
“We believe the FTC needs to look carefully at OpenAI and GPT-4,” said Mark Lautenberg, chairman of CAIDP and a longtime consumer protection advocate on technology issues. rice field.
The complaint attacks various risks associated with generative artificial intelligence since OpenAI’s ChatGPT, which utilizes an early version of the GPT product, first went public late last year. Generative artificial intelligence has captured the world’s attention. Everyday Internet users can use ChatGPT to write poetry, create software, and get answers to their questions. All this within seconds and amazingly sophisticated. Both Microsoft and Google have begun integrating the same types of AI into their search products, and Microsoft’s Bing runs on his GPT technology itself.
But the race for supremacy in a seemingly new field has resulted in unsettling or downright false results, such as the confident claim that February 12, 2023 was before December 16, 2022. is also produced. In industry parlance, mistakes of this kind are known. In its complaint, CAIDP argued that these should be treated as “AI hallucinations” and considered legally enforceable violations.
“Many of the problems associated with GPT-4 are often described as ‘misinformation,’ ‘hallucinations,’ or ‘fabrications.’ But for the FTC’s purposes, these outcomes are best understood as ‘deception,'” the complaint said, adding to the FTC’s broad powers to prosecute unfair or deceptive business conduct and practices. Mentioned.
The complaint acknowledges that OpenAI has been outspoken about many of its algorithm’s limitations. For example, the white paper linked to GPT-4, the latest release of GPT, explains that this model “may generate nonsensical or false content in relation to certain sources.” increase. OpenAI also makes similar disclosures about how tools like GPT can lead to widespread discrimination against minorities and other vulnerable groups.
However, in addition to arguing that these results themselves may be unfair or deceptive, CAIDP argues that OpenAI’s attempts to shift responsibility for these risks onto customers using its technology have led to the FTC’s AI It also claims that it violates guidelines.
According to the complaint, OpenAI’s terms require newspaper publishers, banks, hospitals and other institutions that implement GPT to include waivers about the limitations of artificial intelligence. That doesn’t excuse OpenAI from liability, according to the complaint.
Citing the March FTC advisory on chatbots, CAIDP writes: [the] “Warning customers of abuse or telling them to disclose information is not enough to deter bad actors,” the FTC said. Deterrents should be durable, built-in features, not bug fixes or optional features that third parties may modify or remove to break functionality. ”
CAIDP notes that artificial intelligence can also have a significant impact on consumer privacy and cybersecurity, and while these issues lie directly within the jurisdiction of the FTC, the FTC is linked to the internal workings of the GPT. said he had not studied
