ChatGPT Thinks Americans Are Excited About AI. Most Are Not.


Before you ask: ChatGPT did not write this article. I will admit that I did ask the AI chatbot to write an 800-word article in FiveThirtyEight’s style about current public opinion on AI chatbots like ChatGPT, but our collaboration disintegrated amid editorial differences. (If you’re curious what ChatGPT had to say, though, you can read its take at the bottom of the page.)

Apart from committing the cardinal sin of failing to produce polling numbers until the sixth paragraph, ChatGPT’s take on how Americans see it and other chatbots was suspiciously positive. “A 2021 survey by the Pew Research Center found that 71% of Americans believe it is generally a good thing for society if robots and computers become more capable and sophisticated, while only 27% believe this would be a bad thing,” ChatGPT wrote. “This suggests that people are increasingly open to the idea of AI and see it as a potential source of innovation and progress.”

You should only abandon FiveThirtyEight for chatbot journalism if you don’t mind some made-up data. I was unable to find the 2021 Pew survey that ChatGPT was referencing, and when I reached out, Pew’s media team was similarly stumped about where the question came from. (ChatGPT’s tendency to make up facts is a “core challenge” for the technology, according to its chief technology officer.) In fact, a 2021 Pew survey on the topic that I did unearth suggested the opposite. When asked about the increased use of artificial intelligence in daily life, only 18 percent of respondents said they were more excited than concerned, while 37 percent said they were more concerned than excited and 45 percent said they were equally concerned and excited. This is consistent with what recent polls actually show: Americans generally don’t trust AI — particularly in high-stakes situations, like medical care — and they’re fearful that AI-powered search engines could be inaccurate and biased.

Lest you think I am being too harsh on a technology that’s still in its infancy — it’s not all bad news for ChatGPT. A Morning Consult poll conducted Feb. 11-13 found that a majority (52 percent) of Americans think that generative AI platforms like chatbots are not a fad, and majorities of workers in health care, design, engineering and finance/accounting said that their industry has more to gain from use of generative AI tools than to lose.

But it’s safe to say that AI has a lot of trust to build. According to Morning Consult, only 10 percent of Americans think the output from generative AI is “very trustworthy,” while 11 percent think it is “not at all trustworthy,” and the remaining 80 percent are undecided. Another Pew survey conducted from Dec. 12-18 found that 60 percent of Americans would be uncomfortable if their own health care provider relied on artificial intelligence for their medical care, and 75 percent said they were more concerned that health care providers will move too quickly with this technology, before fully understanding the risks to patients. And speaking of my own profession, a Monmouth University poll conducted Jan. 26-30 found that while 72 percent of Americans believe that there will be a time when entire news articles are written by AI, 78 percent of Americans think this would be a bad thing.

That’s in part because concerns about misinformation, bias and accuracy are fairly widespread. When asked last week by Morning Consult about search engines that rely on AI, roughly two-thirds of adults said they were worried about misinformation, 63 percent said they had at least some concern about accuracy and 62 percent said they were concerned about biased search results. And AI applications like chatbots are sparking more dystopian fears too. According to the survey, 63 percent of adults worry that AI will cause harmful behavior, and a similar share are concerned that AI applications could “learn to function independently from humans.” Those fears seem to be growing: The Monmouth University survey found that 55 percent of Americans now believe that artificial intelligence could pose a risk to the existence of the human race, up from 44 percent in 2015.

So what do Americans think AI is good for? Recipes, roadside assistance and coal mining, according to Morning Consult and Monmouth. The two pollsters approached the question in different ways: Morning Consult’s Feb. 17-19 survey asked respondents if they were “very” or “somewhat” interested in a number of AI-related products and services. AI-powered online search, AI-generated recipes, and AI-powered roadside assistance topped the list, although 43 percent of respondents were “very” or “somewhat” interested in AI tools for police or criminal justice — a use for AI with considerably higher stakes than search accuracy. Monmouth, meanwhile, asked about six potential applications of artificial intelligence and found that respondents were only broadly OK letting AI take on risky jobs like coal mining (75 percent), compared to facial recognition at public places (54 percent), nuclear power plant safety operations (38 percent), or military drone decision-making (34 percent). 

It’s one thing to send a robot down a coal mine. But we’ll see how enthusiastic Americans are about AI-generated recipes when ChatGPT’s culinary equivalent sneaks some horseradish into an apple pie. Based on early forays into AI-fueled recipe generation, that might be a best-case scenario

Other polling bites

  • Ahead of the first anniversary of Russia’s invasion of Ukraine, President Biden made a surprise visit to Kyiv on Monday to show American support for the beleaguered country. A YouGov poll conducted on Feb. 22 found that a slim majority of Americans (51 percent) at least somewhat approve of Biden’s visit. An additional 17 percent neither approve nor disapprove, 22 percent at least somewhat disapprove and 10 percent weren’t sure. This reflects the broader state of public opinion about American support for Ukraine, according to another YouGov poll conducted on Feb. 22: Slightly less than half (46 percent) of Americans say they want the U.S. to “support Ukraine in its efforts against Russia” until Russia withdraws, even if that makes the war last longer, while 24 percent want the U.S. to “encourage a negotiated peace to end the fighting” even if that results in more Russian control within Ukraine, and an additional 30 percent said they wanted neither option or didn’t know.
  • A new survey from Public Religion Research Institute found that the share of Americans who think abortion should be legal in all or most cases has grown since the organization began asking the question, from 55 percent in 2010 to 64 percent today. There’s been even more movement among Democrats: Today, 87 percent of Democrats think abortion should be legal in all or most cases, up from 71 percent in 2010. However, there have not been major shifts on the issue since the Supreme Court overturned Roe v. Wade last summer. The survey, which also looked at views on abortion legality by state, found that there are only seven states where less than half the population thinks abortion should be legal in all or most cases.
  • Most Americans don’t think the country is equipped to handle another pandemic. A new Axios-Ipsos survey on public health found that only 28 percent of Americans believe the country is “adequately prepared” for another pandemic, and 75 percent of Americans don’t think that American businesses or the government prioritize Americans’ health or well-being. That being said, at least 75 percent of Americans said their personal health is “good” on a range of metrics — but when probed, much smaller shares said they’re fully satisfied with their physical health, diet and personal finances.
  • Just compromise, already! That’s what a new NPR/PBS NewsHour/Marist National Poll wants government officials in Washington to do: According to the poll, 70 percent of Americans think it’s more important for politicians to compromise, while 28 percent think it’s more important to stand on principle. However, on the specific issue where compromise is eluding Congress currently, Americans are more divided. Just about half (52 percent) of registered voters support raising the federal debt ceiling. This is, however, a much higher level of support than the pollster found during another debt ceiling showdown in April 2011, when only 24 percent wanted to raise the federal debt ceiling.

Biden Approval

According to FiveThirtyEight’s presidential approval tracker, 43.2 percent of Americans approve of the job Biden is doing as president, while 51.7 percent disapprove (a net approval rating of -8.5 points). At this time last week, 43.1 percent approved and 51.8 percent disapproved (a net approval rating of -8.8 points). One month ago, Biden had an approval rating of 42.2 percent and a disapproval rating of 52.3 percent, for a net approval rating of -10.1 points.

I asked my copy editor to edit the following output from ChatGPT. He flagged issues of factual error, style and grammar, voice and tone, and unsubstantiated claims. The resulting piece, with his comments, is below.

Artificial intelligence chatbots have become increasingly popular in recent years, with more and more companies and individuals using them to automate their customer service, provide personalized recommendations, and even hold human-like conversations. ChatGPT, a large language model developed by OpenAI, is one such chatbot that has garnered significant attention and praise for its ability to understand and generate natural language. However, despite its many impressive capabilities, public opinion on AI chatbots like ChatGPT is mixed.

On one hand, there are those who view AI chatbots as a valuable tool that can improve efficiency and productivity in a variety of industries. For example, businesses can use chatbots to automate routine tasks like answering frequently asked questions, scheduling appointments, and processing payments, freeing up human employees to focus on more complex tasks. In addition, chatbots can provide 24/7 support to customers, reducing the need for human staff to work long or irregular hours.

Furthermore, some people appreciate the convenience and personalization that chatbots can provide. Chatbots can use natural language processing to understand and respond to customer inquiries in real-time, allowing them to receive instant feedback and support. They can also use machine learning algorithms to personalize their responses based on previous interactions with the user, making the experience feel more human-like and tailored to the individual.

On the other hand, there are those who are skeptical of AI chatbots and worry about their potential negative consequences. One concern is that chatbots could replace human workers, leading to job loss and economic inequality. For example, if a company replaces its customer service representatives with chatbots, those employees would be out of work, and the company would save money by not having to pay their salaries and benefits. This could exacerbate existing economic inequalities, as lower-skilled workers are more likely to be replaced by automation than those with more specialized skills.

Another concern is that chatbots may not be able to handle complex or sensitive situations as well as humans. While AI chatbots are trained to respond to a wide range of questions and scenarios, there are limits to their abilities. They may struggle with nuance, cultural differences, and emotional intelligence, which could lead to misunderstandings and miscommunications. Additionally, some users may be uncomfortable with the idea of sharing personal information or discussing sensitive topics with a machine rather than a human.

Despite these concerns, the overall public opinion on AI chatbots seems to be generally positive.
According to a 2021 survey by PwC, 49% of consumers say they are comfortable with AI chatbots handling their customer service inquiries, and 59% say they have used a chatbot for customer service in the past year. This suggests that many people are already familiar with chatbots and see them as a viable option for getting support and information.

Furthermore, there is evidence that people are becoming more accepting of AI in general. A 2021 survey by the Pew Research Center found that 71% of Americans believe it is generally a good thing for society if robots and computers become more capable and sophisticated, while only 27% believe this would be a bad thing. This suggests that
people are increasingly open to the idea of AI and see it as a potential source of innovation and progress.

However, public opinion on AI chatbots may vary depending on the specific use case and industry. For example, a chatbot designed to provide personalized shopping recommendations may be more well-received than a chatbot designed to provide medical advice or mental health support. In addition, the age, gender, and educational background of the user may also impact their perception of chatbots and their willingness to interact with them.

In conclusion, public opinion on AI chatbots like ChatGPT is mixed, with some people viewing them as a valuable tool for improving efficiency and productivity, while others worry about their potential negative consequences.

Leave a Reply

Your email address will not be published. Required fields are marked *