PEW Research Releases New AI Report

Pew Research Center released a report that summarizes an in-depth study they conducted through interviews and surveys. They compared US citizens with AI experts. Link to the report is here. Despite the 107 page pdf download, it is easy to skim and find data highlights.

Too many of the questions are hypothetical and ask consumers to answer questions they aren't capable of answering with accuracy. It's a little like asking me who is going to win the World Series this year. I read the papers and catch a game once a year. That doesn't give me any know or guess with certainty which team might win.

Here is what I found interesting:

• 27% of the US public said that they use AI constantly throughout the day (i.e., many times throughout the day). Feels high to me. I am not even sure we'd have these numbers in the Silicon Valley bubble of early adopters.

• 33% have used an "AI chatbot." Among those who have used an AI chatbot just 33% found it extremely or very useful. It is not entirely clear to me how one would know if they were chatting with an AI-powered chatbot. FWIW, only 21% of the same group claimed that it wasn't useful at all. There is a lot of nuance here with complexity of ask/task, alternatives, etc.

• US public believes that 1) users will get inaccurate information from AI (66%), 2) AI will impersonate people (65%), and 3) there will be a loss of human connection (57%). I have two reactions. First, I'm surprised that consumers are this concerned. These numbers are all likely a bit high - just a nuance with surveys. Second, each of these things is happening and will happen - by design and accident. It worries me that these numbers aren't higher. Folks should know the risks.

• White males are the most likely to feel they are represented well in "AI design." They are probably right.

• Few consumers believe that AI can outperform people whose jobs are to drive someone from one place to another (19%), provide customer service (19%), decide who can get a loan (19%), make a medical diagnosis (26%), write a news story (19%), decide who gets parole from prison (10%), make a hiring decision (11%), or write a song (14%). I don't believe that AI agents will replace humans even if they can do some tasks humans do today. I appreciate the optimism around humans.