Please ensure Javascript is enabled for purposes ofwebsite accessibility

NewsChannel 12 Investigates: Artificial Intelligence Part III


{p}{/p}

Facebook Share IconTwitter Share IconEmail Share Icon

NewsChannel 12 investigates artificial intelligence as it transforms our future.

From exposing AI’s potentially revolutionary capabilities to dangerous disruptions and what lawmakers are up against.

For some it’s difficult to spot the difference between what’s real and fake. The errors from AI may be quite subtle and at times unrecognizable.

“I was previously very skeptical that we would ever reach this level of sophistication with our artificial intelligence tools,” said Dr. Chris Callison-Burch, an associate professor of computer and information science at The University of Pennsylvania. “Although, it feels like an overnight success, we researchers have been at it for decades in order to make it.”

Text, images, audio, and video are generated in seconds with the use of AI.

“We’re just now scratching the surface of what’s possible,” Callison-Burch said.

His team of students have proven AI is still catching up to how human brains work.

“A key thing to keep in mind about these systems is they are not generating text that is fact checked,” he said. “They’re producing the next word in the same way that your cell phone is when it’s trying to autocomplete your sentence.”

Callison-Burch created a training game called “Real or Fake Text.” It includes short stories, news articles, recipes, and presidential speeches.

In each line the participant must guess whether it’s written by a human or machine.

“Each new word they’re generating is based on the previous context,” Callison-Burch said. “As a result, they tend to generate very amazingly grammatical, plausible sounding texts.”

Callison-Burch said if the text doesn’t add up, it’s likely artificial intelligence.

“The types of errors that the computers make tend to be things that we as human beings would consider common sense,” he said.

AI can easily create images too, which may be difficult and dangerous.

Callison-Burch used WhichFaceisReal.com to point out AI images.

“One of which is a photograph of an actual person and the other one is as imagined person that has been generated through an artificial neural network,” he said. “Oftentimes the areas outside the face region are distorted or nonsensical.”

Callison-Burch urges people to watch for small details like fingers, ears, hairlines, and accessories like glasses. It may turn out distorted, which is a sign of artificial intelligence.

Some images may depict dramatic events like one of a protest turned violent in New York. After a close examination it was ruled to be fake.

Images are now easy and sometimes free to make and share.

“It is getting much better and therefore it will be harder for people to spot,” Callison-Burch said. “All members of a democratic society need to be skeptical of the information that they’re being presented with, they need to think about the source and need to rely on credible sources of information.”

Callison-Burch testified in a House of Representatives hearing on AI and copyright law. He told the NewsChannel 12 Investigates Team he hopes the technology outweighs the negatives, including advancements in entertainment and workplace productivity.

Technology Companies Open AI, Microsoft, and Google stated they’re committed to safe and responsible AI.

Watch more NewsChannel 12 investigations into artificial intelligence.

Contact the NewsChannel 12 Investigates Team at (252) 571-9209 or email at iteam@wcti12.com

Loading ...