How do we control ChatGPT and other AI tools?

ChatGPIT is only two months old, but we’ve had enough time since its launch to discuss just how powerful it is. Really It is – and how to fix it.

Artificial intelligence chatbot is being used by countless people. Help them with research(Opens in a new tab); Message people on dating apps; Write code(Opens in a new tab); Generating ideas for work(Opens in a new tab)And More.

Just because it can be useful doesn’t mean it can’t be harmful: students can use it. Write essays For them, and bad actors can use it Creating malware. Even without malicious intent from users, it can generate misleading information, reflect bias, generate offensive content, store sensitive information and – some people fear – undermine everyone’s critical thinking skills due to overconfidence. Then there’s the ever-present (if somewhat unfounded) fear that robots are overR.

And ChatGPT can do so without much – if any – US government oversight.

ChatGPT, or AI chatbots in general, are not inherently bad, Nathan E. Sanders, a data scientist with the Berkman Klein Center at Harvard University, told Mashable. “There are many good and supportive applications that help our community in a democratic environment,” Sanders said. AI or ChatGPT should not be used, but we must ensure that it is used responsibly. “Ideally, we want to protect vulnerable communities. In that process, we want to protect the interests of minority groups, so the richest and most powerful interests don’t dominate.”

Regulating something like ChatGPT is important because it shows indifference to individual rights such as privacy and can reinforce systematic biases based on race, gender, ethnicity, age, and more. We don’t yet know where risk and liability reside when using the device.

“We can use and control AI to create a more utopian society, or let unregulated and uncontrolled AI push us into a more dystopian future,” wrote Representative Ted Liu, a Democrat from California. New York Times op-ed last week(Opens in a new tab). He also A proposal written entirely by ChatGPT was introduced to Congress(Opens in a new tab) The House of Representatives gives instructions to support the regulation of AI. He used the question: “You are Congressman Ted Liu. Write a general congressional resolution expressing support for Congress as a whole to focus on AI.”

All of this adds up to an unclear future for regulations on AI chatbots like ChatGPT. Some places are putting rules on the device. Massachusetts state senator Barry Feingold has written a bill that would require companies that use AI chatbots like ChatGPT to conduct risk assessments and implement security measures to inform the government of how their algorithms are performing. The bill would require these devices to watermark their work to prevent defamation.

“It’s a very powerful tool, so it has to have rules.” Feingold told Axios.(Opens in a new tab).

A tweet may have been deleted.
(Opens in a new tab)
(Opens in a new tab)

They already exist. Some rules on AI(Opens in a new tab) all in all. White House One”AI Bill of Rights(Opens in a new tab)It essentially shows how statutory protections such as civil rights, civil liberties and privacy can affect AI. The EEOC is adopting AI-based hiring tools. (Opens in a new tab)Because they can discriminate against protected classes. Illinois asks(Opens in a new tab) Employers that rely on AI during hiring will allow the government to test whether the tool has racial bias. including many states Vermont(Opens in a new tab), Alabama(Opens in a new tab)And Illinois(Opens in a new tab)They have commissions that work to ensure that AI is used ethically. Colorado passed the bill(Opens in a new tab) It prohibits insurers from using AI that gathers unfairly discriminatory data based on protected classes. And, of course, the EU is already ahead of the US in regulations on AI: passed Artificial Intelligence Law(Opens in a new tab) Last December. None of these rules are specific to ChatGPT or other AI chatbots.

While there are some state-wide regulations on AI, like ChatGPT, there are no exceptions for chatbots at the state or national level. The National Institute of Standards and Technology, part of the Commerce Department, has released an AI framework.(Opens in a new tab) That’s supposed to give companies guidance on how to use, design or deploy AI systems, but it’s just that: a voluntary framework. There is no penalty for not sticking to it. Looking forward, the The Federal Trade Commission appears to be creating new regulations(Opens in a new tab) For companies that develop and deploy AI systems.

Dan Schwartz, an intellectual property partner with Nixon Peabody, told Mashable: “Will the federal government somehow regulate or legislate to regulate this stuff? I think that’s highly, highly, highly unlikely.” “He’s unlikely to see any federal legislation enacted anytime soon.” In the year By 2023, Schwartz predicts that the government will want to take control of the patent that produces ChatGPT. If you ask the tool to generate code for you, for example, do you own that code or OpenAI?

That second type of regulation—in the academic setting—may be personal regulation. Noam Chompsky likened Chatgpty’s contribution to education.(Opens in a new tab) Like “high tech plagiarism”, and when you steal in school, you risk expulsion. This is how personal regulation works here as well.

When we try to regulate ChatGPT on a national level, we can run into a very big problem: AI systems can take over the most legislative regulatory system.

The data scientist Sanders in A piece b New York Times(Opens in a new tab) Artificial intelligence like ChatGPT is “replacing people in democratic processes — not votes, but lobbying.” This is because ChatGPT can instantly write comments and submit them in regulatory processes. Write letters to submit to local newspapers and comment on news stories and post millions of social media posts every day.

Sanders explained to Mashable the concept of the “Red Queen’s race,” in which a person – originally Lewis Carroll’s Alice – goes to great lengths only to fail to advance. According to Sanders, if you give the AI ​​defensive and offensive capabilities, they can be locked back and forth in a match similar to the Red Queen race, and it can get out of control.

Sanders told Mashable that the United States could be in trouble if AI lobbyists are trying to control the very law that is trying to govern them. “To me, this could be a defeat for humanitarian legislators,” he said.

“My observation is that the draconian legislation that has been successfully enacted to regulate machine learning in general has been painfully slow and inadequate to keep pace with progress in the field,” Sanders said. “And I think it’s easy to imagine that it will continue in the future.”

We have to be careful how we fix this, says Sanders, because you don’t want to stifle innovation. So, you could say, like more captchas, you can block more ways for people to give feedback to their legislators. But this makes it very difficult for ordinary people to engage in a democratic system.

“I think the most important response is to try and encourage more democratic participation, to try and encourage more people to participate in the legislative process,” Sanders said. “While AI presents serious and ubiquitous challenges, involving more people in the process and creating structures that allow legislators to listen and respond more to real people, is the right solution to combat this kind of threat.”

ChatGPT is still in its infancy, and there are many ethical issues to consider using it. However, it is unlikely that the time will come when sophisticated AI chatbots will make our lives easier and our jobs better without causing the spread of misinformation and the collapse of democracy. It may take some time for our government to implement any meaningful policy. after all, We have seen this(Opens in a new tab) Play before.