(This is a living document. Link contains latest version of the document.)
I conducted a survey on Prolific.com on whether US citizens support a pause on further AI development. 1484 people responded.
Analysis
Overall more people are against a pause than for, but there is significant mainstream political support for a pause on further AI research in the United States.
Age groups below 25 and 25-34 are most in support of a pause, followed by those above 65.
Women are significantly more in support of a pause than men.
People who are unemployed and part-time employed are significantly more in support of a pause than people with full-time employment.
People of black and mixed ethnicity are most in support of a pause, by significant margin. People of Asian ethnicity are least in support of a pause, by significant margin.
Based on the comments provided by people in the form
Job loss is the number one people reason people support an AI pause.
“Safety concerns” are the second reason, although many people have not been specific what this means to them.
Significant number of people explicitly say they believe in the possibility of AI takeover and support AI pause for this reason.
Significant number of people support a pause if it was conducted worldwide, but do not support US unilaterally pausing because of concerns China would catch up.
I strongly recommend reading the comments in full, there are hundreds of comments on an AI pause, I found a few of them insightful. Full response data is shared below
I also recommend doing your own data analysis and drawing your own conclusions. Full response data is shared below.
Future surveys could include questions on more narrow demographics such as religion, STEM background, political leaning and influential people they defer to on this topic.
Some interesting comments
I have selected a few of the comments from the dataset published below, that I found interesting. This is not a random selection.
Strongly agree—I believe that AI may have some benefits, but the consequences of a rogue AI or even just too much dependence on AI far outweigh the benefits. − 27, Male, White, Part-time employed
Strongly disagree—I think AI is a necessary evil, the further we develop it we’ll see the next few groundbreaking steps, but I believe AI is necessary, even if it does progress in the likes of Terminator. − 24, Male, Mixed
Strongly disagree—A pause won’t stop anything because other regimes would continue. collective action problem − 38, Male, Other, Full-time employed
Strongly disagree—You won’t get every country to agree to a pause, and even if they do, some might continue in secret. Whomever reaches AGI or ASI first is going to have a significant impact on and advantage in the future. We don’t want that to be an adversarial nation. Also, less dramatically, AI provides the opportunity to shape a post-labor economy where people don’t have to work anymore. There will be growing pains, but the future could be utopian. − 46, Male, White, Part-time employed, US
Strongly disagree—I don’t think that AI will destroy the human race. I also think that the videos are deepfakes. − 60, Male, Black
Strongly agree—AI is going to wake up and destroy this planet and everybody on it − 59, Male, Black
Strongly disagree—I believe that AI should continue because it is helping many people. − 26, Female, Black
Strongly agree—I think that AI is getting out of hand and dangerous, especially with AI videos that are indistinguishable from reality. If we don’t stop allowing this technology to evolve, it could get even more dangerous. The common sentiment I have been seeing online is that majority of people are sick of this. − 22, Female, White
Weakly agree—I do not agree with the AI statement linked above or think that AI poses an “extinction risk,” but the rate at which datacenters are being expanded to power LLMs considerably contributes to climate change, which IS an existential risk. So I support the pause for environmental reasons. − 30, Female, White, Student and full-time employed
Strongly agree—Why would someone program an AI program to kill all humans? The programmers are human too! − 45, Female, White
Weakly disagree—i feel that man always think they are the smartest but God in Heaven will have the final say. Self destruction because one is too many and one thousand is never enough. Greed, power and the things of the world has led to this destruction. It has to play out now. − 56, Female, Black
Weakly agree—I only weakly agree because of the issues AI is causing in poor and rural communities. The usage is killing members of those communities. − 34, Female, Mixed
Weakly agree—I am not sure because I love how AI helps me with every aspect of my life and has drastically changed the quality of my life but I need to be more educated on what the overall goals and intentions are of AI so really at this point I need to gather more information. My partner’s aunt recommended us this book on the subject. − 43, Female, Mixed, Full-time employed
Weakly agree—indifferent − 22, Female, White
Strongly agree—I believe that the environmental impacts of AI are too severe to keep ruthlessly expanding it and that we need to take time to build the infrastructure to support it IF it were to continue. − 21, Female, White
Weakly Agree—I disregard anything Musk has to say bc he’s a fraud, so any video featuring him and his insight makes me skeptical of everyone in it. His only knowledgeable lies in spreading propaganda to morons and running businesses into the ground. That being said, I do sometimes think of the possibilities. Mostly bc my husband has been saying this ever since the Terminator movies came out and it can be terrifying to think about. I think we need to put the brakes on a bit and get some really strong regulations and oversight. Apologies for the rant. − 53, Female, White, Other employment
Strongly agree—The videos had the aha effect on me. I never thought about the negative implications of AI in terms of harming mankind with superintelligence. It’s pretty scary. − 70, Female, White, Not in paid work
Strongly disagree—If we are still at a point where humans are in control, then humans should create a limit...a threshold...for AI to not be able to surpass. That is, “if” we are still at the point of being in control. I don’t know if we are or if we are not. − 66, Female, White
Weakly disagree—The loudest voices in this fear of AI are the same people who think the government is nefarious and that COVID was a hoax. Many others seem to be against it because of a competing agenda. I think the concerns are overblown. − 55, Female, White
Neither agree nor disagree / No opinion—i think the way the people in those videos think. It’s scary to think about and possible that AI will take over − 46, Female, Mixed, Not in paid work
Strongly agree—I think there’s enough predictions from people who know more about AI that are not necessarily positive outcomes for humanity. − 45, Female, Mixed, Unemployed
Weakly agree—I don’t think the general public is given enough information on the potential damage, including environmental, caused by furthering AI and I especially do not think the current US government is capable of meaningful strategic planning and effective federal controls. − 61, Female, White (German origin)
Weakly disagree—You would just let China take over. − 68, Male, White (Cuba origin), Other employment
Weakly agree—More developments of AI can lead to human extinction − 30, Male, Black (Kenya origin)
Strongly agree—I support but need to have clear policy to use. − 41, Female, Asian (Nepal origin)
Weakly disagre—AI has its pros and cons just like everything else. I am open to AI but know it can also do more harm than good in some cases and vice versa. − 43, Female, Asian (Korea origin)
Strongly agree—There are many other scientific ventured, such as medicine, global warming, food, clean water etc that need more widespread support than AI research does. AI is statistically horrible for the planet. − 28, Female, Asian (India origin), Unemployed
Weakly disagree—I believe a global pause on AI research is difficult to enforce at this point. From governments to companies to everyday social media users, AI is at people’s fingertips — and as such someone somewhere will certainly want to study this theme. Perhaps a shift in perspective to better protocols and controls for AI development is a stronger alternative. − 26, Male, Other (Peru origin)
Data privacy and consent
Data that will NOT be published publicly: Name, Prolific ID, Personally identifying info (email ID, phone, social media ID, home address)
Data that will be published publicly: Responses to survey below, Fluent languages, country of birth/residence/citizenship, age, sex, ethnicity, student/employment status
[ ] Yes, I consent
I support a pause on further AI research
[ ] Strongly agree
[ ] Weakly agree
[ ] Neither agree nor disagree / No opinion on this topic
[ ] Weakly disagree
[ ] Strongly disagree
Short answer question: Prolific ID (required)
Paragraph answer question: What are your thoughts on this topic? (OPTIONAL)
Prolific.com survey on AI pause
Link post
(This is a living document. Link contains latest version of the document.)
I conducted a survey on Prolific.com on whether US citizens support a pause on further AI development. 1484 people responded.
Analysis
Overall more people are against a pause than for, but there is significant mainstream political support for a pause on further AI research in the United States.
Age groups below 25 and 25-34 are most in support of a pause, followed by those above 65.
Women are significantly more in support of a pause than men.
People who are unemployed and part-time employed are significantly more in support of a pause than people with full-time employment.
People of black and mixed ethnicity are most in support of a pause, by significant margin. People of Asian ethnicity are least in support of a pause, by significant margin.
Based on the comments provided by people in the form
Job loss is the number one people reason people support an AI pause.
“Safety concerns” are the second reason, although many people have not been specific what this means to them.
Significant number of people explicitly say they believe in the possibility of AI takeover and support AI pause for this reason.
Significant number of people support a pause if it was conducted worldwide, but do not support US unilaterally pausing because of concerns China would catch up.
I strongly recommend reading the comments in full, there are hundreds of comments on an AI pause, I found a few of them insightful. Full response data is shared below
I also recommend doing your own data analysis and drawing your own conclusions. Full response data is shared below.
Future surveys could include questions on more narrow demographics such as religion, STEM background, political leaning and influential people they defer to on this topic.
Some interesting comments
I have selected a few of the comments from the dataset published below, that I found interesting. This is not a random selection.
Raw data
Link to full response data
(Explicit consent was obtained via google form before publishing this.)
Prolific.com study details (filled before starting survey)
Study Name: Do you support a pause on further AI research? (Single question survey)
Study Description: Only one question in this survey: Do you support a pause on further AI research?
Data collection type: Survey
Which devices can participants use to take your survey: Mobile, Tablet, Desktop
Does your study require any of the following: No (No Audio, No Camera, No Mic, No Software Download)
Content warning: None
What’s the URL of your study: (link to google form here)
Does your study have a limit of participants that can enter at the same time?: Don’t limit participant access (recommended)
Recording Prolific IDs, How do you want to record IDs?: A question in my study
Custom screening: No
Completion paths, How do you want to process submissions?: Automatically approve the submission and pay the participant
Completion paths, redirect URL: (this code and URL was inserted into the completion of the google form)
Recruit participants: Find new participants on Prolific
Participants: 1500 (study was later stopped before 1500 participants, as finding old mixed and other males was taking too long)
Credentials required: No
Location (of participants): USA
Study distribution: Representative sample (Distribute your study based on UK or USA census data)
Representative sample criteria: USA, Factors: Sex, Age, Ethnicity (Simplified US Census)
Estimated time to complete survey: 1 minute
Payment: 0.10 GBP per participant
Total cost: 200 GBP (150 GBP to participants, 50 GBP to Prolific.com)
(~10 responses out of ~1500 were rejected due to minor issues with the approval system)
Prolific.com study details (after completing survey)
Eligible participants: 63773 of 176424
Median time (to complete survey): 1 minute 22 seconds
Google form details (created before starting survey)
Title:
Do you support a pause on further AI research?Text:
(About video)
(Video title: AI Experts on extinction risk)
(Video Length: ~1 min)
(Video previw opens in google form, no need to visit youtube to watch the video.)
Text:
(About video)
(Video title: AI Experts on extinction risk)
(Video Length: ~1.5 min)
(Video previw opens in google form, no need to visit youtube to watch the video.)
Text:
Checkbox question: (required)
Multiple choice (single answer) question (required)
Short answer question:
Prolific ID(required)Paragraph answer question:
What are your thoughts on this topic? (OPTIONAL)