Security camera footage from outside a bar in the Seattle area in 2021 captured a tragic event where three individuals lost their lives. The screenshot of the footage was obtained from KOMO News (Fair Use).
A recent development in Washington state involves a judge who has prohibited the submission of video evidence that has undergone “AI-enhancement” in a triple murder trial. This decision is significant as it sheds light on the misconception that applying an AI filter can reveal hidden visual data.
Judge Leroy McCullough in King County, Washington, highlighted concerns about the use of AI technology, stating that the methods used by AI models to determine what should be displayed are not transparent. This ruling brings clarity to the potential issues surrounding the use of AI tools amidst the prevalent hype surrounding artificial intelligence.
The case revolves around Joshua Puloka, a 46-year-old individual accused of the murder of three individuals and the injury of two others at a bar near Seattle in 2021. Puloka’s defense team sought to introduce cellphone footage captured by a bystander that had been enhanced using AI technology. However, the rationale behind presenting the altered footage remains unclear.
Reportedly, Puloka’s lawyers enlisted an “expert” in creative video production with no prior experience in criminal cases to enhance the video. The AI tool utilized by this unnamed expert was developed by Topaz Labs, a Texas-based company accessible to the general public.
The proliferation of AI-powered imaging tools has led to widespread misunderstandings regarding the capabilities of such technology. Contrary to popular belief, AI upscalers do not enhance the existing visual information in an image or video; instead, they introduce new information that was not originally present.
An example cited is the debunked conspiracy theory surrounding Chris Rock wearing a face pad during the incident where he was slapped by Will Smith at the 2022 Academy Awards. The theory emerged from individuals using AI upscalers on screenshots of the incident, hoping to uncover hidden details. However, the enhanced images were merely distorted representations, not accurate depictions of reality.
The emergence of AI-labeled products has fueled confusion among the general public about the true capabilities of these tools. Large language models like ChatGPT, often perceived as capable of complex reasoning, primarily function by predicting the next word to emulate human-like responses. Despite their linguistic proficiency, these models do not engage in sophisticated reasoning but rather mimic human conversation effectively.
The judge’s decision in Washington underscores the limitations of AI technology in enhancing visual content. While some judges may be swayed by the AI hype, it is crucial to recognize the inherent shortcomings of these technologies. As investments in AI companies continue to rise, it is essential to maintain a critical perspective on the actual capabilities of AI tools to avoid misconceptions and misplaced trust in their functionality.
