Tuesdata: How the public started paying attention to AI, but don’t yet see the benefits
Welcome to Tuesdata, for Unmade’s paying members.
In today’s post we explore what the data tells us about public attitudes to the rise of generative AI. Those inside the communications bubble are both excited and concerned, but are members of the public even paying attention yet? The data suggests they are – and Australians in particular are among the most ambivalent.
If you’re not a paying member, you’ll hit a paywall further down. Below the wall, you’ll also find the discount code to save an extra 30% on tickets to humAIn – human creativity x AI.
How generative AI slipped into the the mainstream
It’s only four months since the release of OpenAI’s ChatGPT radicalised expectations for the impact that generative artificial intelligence was about to have on the business world.
Those within the media and marketing industry have already started to focus on how the new technology will change the way they do business. But how about the public as a whole?
We’ve turned to Google Trends, and an in depth study of public opinion across the world, to understand how the world is waking up to AI in general and ChatGPT in particular.
First, to ChatGPT. The prototype was launched on November 30. Based on Open AI’s large language model Generative Pre-trained Transformer 3, ChatGPT instantly became the most widely adopted and talked about chatbot in history, reaching a million users within its first week.
Not only did search interest explode, but Chat GPT quickly overtook the general term “AI” for people interested in the topic.
The moment when ChatGPT moved past AI as a search term occurred around January 15. By contrast, it took Google a couple of years to become a verb for ‘search’, back at the end of the last century.
A similar battle of the brands has been occurring in the field of AI-driven image generation in what has mainly become a two-way fight between DALL-E, powered by OpenAI, and Midjourney.
DALL-E was first to market. But Midjourney quickly overtook it in terms of public search interest. One early advantage enjoyed by DALL-E was the ability to use it via a simple login, while Midjourney requires the more convoluted process of joining and using Discord to generate images. However, Midjourney’s images tend to be more impressive.
And the rise of image-based generative AI brings with it concerns of fake images which purport to show true events.
An intriguing element of Donald Trump’s criminal indictment this month was the proliferation of AI-generated images of Trump being arrested by police. The viral images seem realistic enough at first glance – but they were made with Midjourney.
An image of the Pope apparently wearing a Balenciaga puffer jacket went similarly viral.
The trust gap is real. A study by KPMG Australia and the University of Queensland on attitudes towards AI suggested that Australians were the least convinced that things like enhancing how decisions could be made, costs could be saved or outcomes could be improved would have a genuine benefit.
Of the 17 countries covered, Australia was bottom.
However, the study also seems to suggest that Australians are among the least concerned about the implications. They were among the least outraged and fearful.
Some 42% of Gen X and Millennials said they trusted AI, compared to just 25% of older generation Australians.
Asked specifically about perceived risks, Australian consumers sat in the middle of the pack.
Perhaps because of the high profile Optus and Medibank hacks, (which were not conducted by AI) Australians are among the populations most concerned about cyber crime.
Another factor which could have implications for brands is the emergence of AI-created faces judged as more trustworthy than the real thing. This contradicts the “Uncanny Valley” theory that humans can instinctively identify, and be repulsed by replicas.
A collaborative study conducted by Lancaster University in the UK and the University of California in the US asked participants to identify a selection of 800 faces as real or fake, and to rate their trustworthiness.
After three separate experiments, the AI-created synthetic faces were found to have on average rated 7.7% more trustworthy than the average rating for real faces.
The levels of trust evoked by the synthetic faces was broadly and markedly higher, with the top three most trustworthy faces being synthetic, while the bottom four least trustworthy faces were real.
High trust faces – which boast features like wide set eyes, U-shaped mouths, higher amounts of facial fat, high inner eyebrows and pronounced cheekbones – could easily be generated using tools like Midjourney and deployed in advertising material.
Public opinion – and knowledge of AI – is changing fast. Repeat all of the above studies in another year, and the results could be entirely different.
Time to leave you to your Tuesday.
Thank you as always for supporting Unmade through your membership.
Don’t forget to take advantage of your extra 30% discount on top of the earlybird price for humAIn – human creativity x AI. Your coupon code is below. Tickets are now on sale.
Your coupon code for humAIn
Your humAIn discount code is:
Unmade_Member
Go to humain.au and enter the coupon code to access your discounted tickets
