Spot the Bot
Learning digital literacy in an AI world
If I were to show you a video, would you be able to spot if it’s generative artificial intelligence (AI)?
Even the most tech-savvy people are having a harder time catching AI unless it’s labelled. With growing online presence and the approaching Edmonton municipal election, how can people identify when something is generative AI? And is there anything good coming out of AI models?
Generative AI creates new content, like videos, text or photos, while other AI models are reactive or predictive.
“We're in [an] AI bubble right now,” says Matthew Guzdial, computing science associate professor at the University of Alberta and a Canada CIFAR AI Chair with the Alberta Machine Intelligence Institute. “We’re unfortunately in a situation where the legislation and control around these things isn't really keeping up with industry.”
The Canadian Government’s most recent development in AI legislation is a strategy and legal framework. There is also a spot AI content resource on the Canadian government website. Ultimately, AI content is developing faster than governments can regulate, so it’s up to the average citizen to become AI literate.
Catching AI
Ultimately, it’s best to know your source.
According to Guzdial, the best way to avoid being ‘fooled’ by AI is “trusting the people that you know and really recognizing your local community.”
If you know how the person looks, how they write, then you can usually spot if it’s AI-generated. Exercise caution and look over most things across your online spheres with a critical eye. Not everything being shared online is factual.
For photos and videos, AI sometimes generates extra body parts like extra fingers. People in these videos look overly smooth, even plastic-like. If it looks too perfect to be real, it probably is. The other catch is proportions:check if someone is partially blocked, and if the rest of their body makes sense. AI also loves hard contrast for lighting, and Guzdial explains that objects will likely have an “aura of light around it, as if somebody is being lit from all possible angles.” However, it can still be difficult to detect.
AI text is difficult to identify. AI creates its own sources, or sources nonfactual information, so you can check any external links. There are also AI detectors like Copyleaks, QuillBot, Sapling Undetectable AI, and wordtune, but Guzdial cautions against text automated detectors because it’s easy to get false positives.
Many news sites will have AI use guidelines for their journalists, so it’s important to check what their rules are. For example, CBC will never publish anything without human oversight. Postmedia, a corporation that runs local newspapers such as the Edmonton Journal, the Edmonton Sun, Sherwood Park News, Beaumont News, and many other small newspapers across Canada, has AI guidelines.
AI Resources
AI can be difficult to identify for misinformation, but it also has positive uses and can streamline different processes.
Free sources like Khan Academy help to teach AI to students and how to approach it. There are also countless free AI tools that can help anyone create a photo, video, and text. In an Amii Podcast episode, Guzidial mentions that one of his goals is to create an AI program that can allow someone without coding knowledge to create their own personal video games, similar to how people can create their own movies.
AI is also used in the medical field. Guzdial is working with other researchers to use AI for predicting pre-term birth risk and organ transplant failure risk, with an AI model trained on a single laptop.
Future of AI
The AI growth we see today is not going to stay the same. “That bubble is going to pop at some point,” says Guzdial.
Guzdial explains that AI has a summer and a winter. AI summer brings lots of funding and interest, while AI winter is where all funding goes away.
“I'm very concerned because the reason why we had AI winters in the past is because of people overhyping what is possible with AI,” says Guzdial. “There's really important and good things we could do with this technology, [we] just unfortunately have bad actors that are overhyping it right now.”
The free tools will eventually become expensive and investors are going to dwindle, but some AI tools will remain.
“We're not going to see these massive investments that we do right now, but I do think that we're going to see real things that are really benefiting people's day-to-day lives.”
Thank you for alerting us to AI fakery but also it’s positive uses in an accessible article. Rat Creek Press, Edmonton’s free newspaper, addresses a wide range of articles by real human being writers!