16 Great Quotes about AI Fakes & Detection

There are limited guardrails to deter politicians and their allies from using AI to dupe voters, and enforcers are rarely a match for fakes that can spread quickly across social media or in group chats. The democratization of AI means it’s up to individuals — not regulators — to make ethical choices to stave off AI-induced election chaos. – Washington Post

Adobe surveyed more than 2,000 people in the U.S. and 63% of said they would be less likely to vote for someone who uses GenAI in their promotional content during an election. – Fast Company

Even a false-positive rate in the single digits will, at the scale of a modern social network, make tens of thousands of false accusations each day, eroding faith in the detector itself. - IEEE Spectrum 

It took me two days, $105 and no expertise whatsoever to launch a fully automated, AI-generated local news site capable of publishing thousands of articles a day—with the partisan news coverage framing of my choice, nearly all rewritten without credit from legitimate news sources. I created a website specifically designed to support one political candidate against another in a real race for the U.S. Senate. And I made it all happen in a matter of hours.- Wall Street Journal

"Tools to detect AI-written content are notoriously unreliable and have resulted in what students say are false accusations of cheating and failing grades. OpenAI unveiled an AI-detection tool in Jan, but quietly scrapped it due to its “low rate of accuracy.” One of the most prominent tools to detect AI-written text, created by plagiarism detection company Turnitin.com, frequently flagged human writing as AI-generated, according to a Washington Post examination." – Washington Post

It’s important to remember that generative models shouldn’t be treated as a source of truth or factual knowledge. They surely can answer some questions correctly, but this is not what they are designed and trained for. It would be like using a racehorse to haul cargo: it’s possible, but not its intended purpose … Generative AI models are designed and trained to hallucinate, so hallucinations are a common product of any generative model … The job of a generative model is to generate data that is realistic or distributionally equivalent to the training data, yet different from actual data used for training. - InsideBigData

“No single tool is considered fully reliable yet for the general public to detect deepfake audio. A combined approach using multiple detection methods is what I will advise at this stage." Politifact

Too many educators think AI detectors are ‘a silver bullet and can help them do the difficult work of identifying possible academic misconduct.’ My favorite example of just how imperfect they can be: A detector called GPTZero claimed the US Constitution was written by AI. – Washington Post

Most deepfake audio detection providers “claim their tools are over 90% accurate at differentiating between real audio and AI-generated audio.” An NPR test of 84 clips revealed that the detection software often failed to identify AI-generated clips, or misidentified real voices as AI-generated, or both.”  - NPR 

In a year when billions of people worldwide are set to vote in elections, AI researcher Oren Etzioni continues to paint a bleak picture of what lies ahead. “I’m terrified. There is a very good chance we are going to see a tsunami of misinformation.” – New York Times

Google appears to have quietly struck a deal with one of the most controversial companies using AI to produce content online: AdVon Commerce, the contractor linked to Sports Illustrated's explosive AI scandal.  Google is trying to have it both ways: modifying its algorithms to suppress AI sludge while actively supporting attempts to create vastly more of it. – Futurism

Most online detection tools do not provide sufficient information about their development, making it difficult to evaluate and trust the detector results and their significance. - Global Investigative Journalism Network 

Run some of your other writing dated before the arrival of ChatGPT in the fall of 2022 through an AI detector, to see whether any of it gets flagged. If it does, the problem is clearly the detector, not the writing. (It’s a little aggressive, but one student told me he did the same with his instructor’s own writing to make the point.) – Washington Post

Men are quite confident (72%) in their ability to tell real news from fake news than women (59%), according to new polling from the Ipsos Consumer Tracker. We see a similar gender gap when it comes to our perceived ability to tell content that was created by AI. - Ipsos

A former high school athletic director was arrested after allegedly using AI to impersonate the school principal in a recording that included racist and antisemitic comments. The principal was temporarily removed from the school, and waves of hate-filled messages circulated on social media, while the school received numerous phone calls. – CBS News

Dubbed “model disgorgement,” AWS researchers have been experimenting with different computational methods to try and remove data that might lead to bias, toxicity, data privacy, or copyright infringement. – Semafor