AI & the Media: Q&A with Dan Pacheco, Peter A. Horvitz Endowed Chair in Journalism Innovation, USA.

In preparation for an in-depth report on AI and the media, we spoke to several experts in the field via email. One such expert is Dan Pacheco, Peter A. Horvitz Endowed Chair in Journalism Innovation, S.I. Newhouse School of Public Communications, Syracuse University, USA.

What are your current most common use cases of AI?

 I used AI extensively in the production of my upcoming book, primarily for cleaning up transcripts of interviews that I recorded in Zoom. I used Otter.ai to automate the text transcript, then put that into the OpenAI Playground (ChatGPT was not available yet) to format line breaks correctly. I also had it break long sections of text into logically connected paragraphs. This worked very well, but I did have to closely read and re-read every interview because I did find some cases where GTP-3 filled in “holes” in conversations with fabricated information and even artificial dialogue. 

 

 I also used Midjourney to create a book cover. I put the title, chapter titles and a few other keywords Midjourney’s Discord chat and shared what it create with a small group of people who understood the themes of the book. We collectively modified our prompts and then voted on which covers we liked best. The cover that’s up on the Routledge listing was the result. We call it “The Dreaming Eye.”

 

 Here is the prompt that created that image: “/imagine change theories open source XR metaverse AR glasses field test 360 camera data ai automation bots the future”.

 

How else will you use AI in the coming year?

 As a professor of journalism and emerging media technologies, I am actively experimenting with having ChatGPT do the following and am incorporating this into my classes — while also devising methods to ensure that students are still mastering their understanding of the material to do unaided:

• Analyzing data sets given to AI as CSVs and asking it to help find facts and trends for stories. I will have students check the AI’s answers by performing pivot table analyses, sorting and filtering, and so forth in Excel.
• Writing code for web site templates, web-based AR and VR, and functions for interactive environments in Unity and Unreal Engine to help speed up production of immersive story creation.
• Creating imagery for feature stories, or experimenting with ideas that real artists complete.

 

What excites you most about the longer-term future of AI, and why?

Much as the Internet provided easier access to information and learning, I believe that AI will greatly expand our capabilities and efficiency in all kinds of rote tasks and that we will be freed up to perform more creative endeavors.

For example, as a professor I make some assignments that include rubrics which simply check to see that students completed certain tasks. This can become very unwieldy when grading 40 assignments, but I can see myself asking an AI grade helper to tally up the rubrics so that I can spend more time reviewing the more nuanced work that students did, and provide them more feedback that is based on my understanding and experience.

 

 In everyday life, I envision getting helpful suggestions through earbuds or smartglasses on things like traffic or context about what’s happening around me — like a Siri or Alexa that is more of a true personal assistant. Maybe it will even help me remember peoples’ names by recognizing their faces and putting virtual nametags on everyone around me. Making this kind of annotated reality natural and not obnoxious will require us to converse with our AI assistants more like we would our human assistants and colleagues, and I think that’s a good thing. I know I am personally tired of asking Siri on my Apple Watch to be quiet because it misinterpreted a cue.

What concerns you most about the longer-term future of AI, and why?

 

 Unlike many of my colleagues, I am not concerned about AI resulting in widespread cheating — at least in my classes — because I am already adding more checks for understanding that can’t currently be faked with AI.

 

 The bigger thing I worry about is that, just as search engines and mobile devices make it possible for us to outsource memory like phone numbers and facts to technology, AI will make some people even lazier. That said, Plato famously worried that the written word would make it harder for people to remember things, and he was right about that. But isn’t it better to be able to read Homer’s Odyssey instead of memorizing and repeating the entire tale across generations?

 

 I do think that we will need to make ourselves work to not become so dependent on AI and technology that we can’t survive without it. Taking time out to just live and breathe biologically without technology to help us will become an even more important life skill over time thanks to AI.

What are some of the best AI tools (already available for use) that you can recommend to media companies to investigate immediately?

Here are a few:

OpenAI Playground (text): Allows you to paste chunks of textual information into a form. Write what you want the AI to do with it in the first line.

OpenAI GPT-3 Chat (text): Chat with the AI and ask it to do things. Note that you can keep asking it to refine what it sends back to you in a conversation, as if you’re guiding an intern.

OpenAI DALL-E-2 (images): Chat with the AI to create images.

Midjourney (images): Chat with the AI to create images. Requires a Discord account. Start with their instructions on how to create you first image.

AIVA (music): Writes musical scores that begin with genre-centered templates. Once the score is created, you can edit it in a Pro Tools-like interface. (Note, this appeared on this list of 7 other music generators on the Unite.ai site:

Stable Diffusion (text): A latent text-to-image diffusion model capable of generating photo-realistic images given any text input.