One of the various deep fakes Michael Grabowski, Ph.D., developed through ChatGPT’s DALL •E program.
MICHAEL GRABOWKSI / COURTESY
By Mary Haley, asst. features editor and social media editor
BronxNet recently collaborated with Manhattan College professor Michael Grabowksi, Ph.D., on a segment called BronxTalk, regarding the advancements and downfalls of artificial intelligence (AI).
BronxNet is a community station that has been streamed out to all homes in the Bronx since 1993. Many community events, issues and politics that deal with the Bronx are discussed on the channel, while also involving an online presence that provides further outreach.
Manhattan College has developed a close bond with BronxNet over the years, especially with ties between BronxNet and the college’s communication department. Multiple communication students have participated in internships at BronxNet working in their television studios and covering community events with producers and reporters.
MC has also aired some of its programs on BronxNet, including last spring’s panel on how AI has been integrated into various career fields.
As director of the game design and production major and professor in the communication department, AI is something that has concerned Grabowski since its inception, and its more recent takeoff in late 2022 when ChatGPT was introduced. He has brought much discussion on AI to the college’s campus with events like last spring’s panel on AI development in professions such as law, engineering and finance.
“I think the potential of this technology to affect so many disciplines, industries [and] people’s everyday work is pretty vast,” Grabowski said. “I think AI will enhance tasks. There is a human element where AI has a real difficult time replicating, and a lot of creative pursuits AI can sort of mimic, but not really replicate.”
A main point brought up in the segment was the issue of deep fakes. Deep fakes are the ability of AI to produce fake pictures or manipulate images and videos, usually featuring media based on events that have not occurred.
Grabowski brought up the example of an AI developed photo of children in a steam powered factory, working on manufacturing cell phones. The image looked like a real photo taken in the early nineteenth century, which is what Grabowski prompted the AI to render.
Grabowski explained how as you use an AI, in this case ChatGPT’s DALL-E image generator, the algorithm is fed and uses this human interaction to understand how to better itself for future use.
When Grabowski asked the program to give him an image of a drawing done by a six-year-old, he found two main flaws: the “race problem,” where it gave him a white family after he did not mention any race in his prompt, and how it did not understand the realistic capability of the artistic skills of a six-year-old, as it made the image much too sophisticated. After he raised these issues to the AI and asked for another image, it gave him something much more realistic for a six-year-old.
“You can coach the AI and provide it with feedback, and say ‘this is wrong,’ or ‘I want more of this,’ and you’re actually influencing the algorithm as you do that,” Grabowski said. “It’s learning as you are conversing with it.”
Ryan Miller, a professor at Lehman College in the Bronx, explained how with its rapid development we will not be able to get an immediate handle on AI, but this is not an issue we have not seen before.
“If you see any other technology in history…we’ve never controlled the internet, we can’t control the phone network and keep all the scammers from making phone calls,” Miller said. “I don’t think AI is going to be the first time we ever as a society say, ‘Oh, this time we have solved [this] using this technology’.”
The segment ended with a discussion on the politics of this new technology, from AI campaign phone calls being made, which has been a tactic even before recent AI advancements, and deep fakes in the media during election season.
“Everyone needs to have a little bit of media literacy,” Grabowski said. “[Everyone needs to] understand how images are produced.”
Maddy Radion, a game design and computer science double major, hopes to someday use her degree to work for a video game company. With rapid advancements in AI, she spoke to The Quadrangle about how she thinks it might affect her future career.
“I’m not nervous. I’m intrigued,” Radion said. “I found it very interesting that some developers have been using AI to further games in an interesting way…but it can’t replace the real artists and actual workers.”