Over the past year, the rise of AI capable of generating text and images in the last year has provoked doubt and inspired optimism. The journalists who attended our course were no exception.

Journalists are feeling overwhelmed by AI – torn between pushing their bosses to experiment with the new technology so they don’t fall behind, but at the same time afraid the automation enabled by AI will eliminate their own jobs. Other journalists report incorporating AI tools into their workflow: “The cat is out of the bag.”

Those were just some of the findings in a survey of nearly 1,400 journalists who enrolled in a free online course on generative AI and journalism led by Aimee Rinehart, Senior Product Manager AI Strategy for The Associated Press, and me, Sil Hamilton, an AI researcher and Hacks/Hackers Researcher-in-Residence.

Of the 1,380 journalists we surveyed, more than a third thought generative AI will impact them in the short-term, feeling the rapid growth of the AI industry in the last year is overwhelming. Some said they will choose to avoid tools like ChatGPT, while others will explore generative technologies “to better understand how to position” themselves in a new economy. But most of all, respondents wanted newsroom leaders to take the technology seriously. The technology is here, but how to use it is unclear, our students reported.

Aimee and I conducted the survey at the conclusion of our four-week long course, How to use ChatGPT and other generative AI tools in your newsrooms, which was offered by Knight Center for Journalism in the Americas at the University of Texas, Austin in fall 2023.

The course was attended by over 10,600 working journalists from over 147 countries including Nigeria, Egypt, Pakistan, Togo, Ecuador, Japan, Brazil, and more. Afterwards, 1,300 course participants responded to a survey and shared what they thought the implications AI poses for journalism.

Many of the journalists who replied expressed fear that the rapid development and adoption of generative technologies in journalism will leave them overwhelmed and lost. Others were anxious that newsroom leaders aren’t doing enough to engage with what will inevitably become a part of daily life – whether engaging means writing policy to inform newsroom practices or implementing new tools. The survey replies brought to the surface an undercurrent of anxiety: over how to use the technology, over how to avoid it, and over what their professions will look like in the coming years.

“I think more newsrooms and news orgs should be taking the time to do [write AI policies], to engage seriously with the promise and peril of AI,” said one course participant, in a typical response expressing the desire for newsroom leaders to at least engage with AI. Other journalists attending the course and responding to the survey admitted to using tools like ChatGPT and Bard regardless of workplace policy. “The cat is out of the bag.”

Generative systems for advertising have become in vogue over the last year, and many participants saw little preventing these systems from encroaching on the news media landscape. Fear of automation is omnipresent. “This seems like a dangerous path to go down for newsrooms – using AI as an excuse to pile more work on journalists and expect them to crank out more content faster,” predicted another student.

Many participants were frustrated that resources on generative technologies remain fairly technical – that generative AI remains under active research does not help. In turn, while many students appreciated that the course gave an overview of language models, there were many responses expressing a desire to learn more about the new technologies. “Media literacy [classes] in schools (or even for adults) should include sections on [using] Gen AI in journalism,” was a common survey response.

Above all, course participants desired a space to discuss generative AI with journalism colleagues. Many noted the pace of development in the tech industry meant news organizations would need to make a concerted effort to stay relevant and informed. Others felt the monolingual nature of most language models released so far was limiting its potential. Speakers of languages such as Urdu and Spanish were doubtful of applications that rely on natural language processing like ChatGPT would be able to do a good job handling content from their linguistic communities. These survey respondents urged companies like OpenAI to improve their multilingual offerings — especially in light of how useful and ubiquitous these tools are going to become.

The course also featured Q&As with experts and industry leaders like Northwestern University’s Nick Diakopoulos, New York University’s Meredith Broussard, German Public Broadcaster Bayerischer Rundfunk’s Uli Köppen, and AI start-up Steamship, Inc.’s Ted Benson.

The collective comments of the guest speakers indicated a need for flexibility when exploring what generative AI has proven it can offer — while keeping in mind the technology is still immature and prone to problems for now. Powerful new generative AI applications that will transform newsrooms such as ChatGPT are only a year old, but the tech industry has moved quickly in a short period of time.

Understanding how journalism will make use of generative technologies requires spaces to play, explore, and to hack. In response, Hacks/Hackers launched a second course with Sil Hamilton and the Knight Center entitled “Generative AI For Journalists,” that will conclude by the end of December 2023. Hacks/Hackers looks forward to presenting further resources in the coming year.