Generative AI’s application to brand safety is said to speed up the process and be more accurate

Pixability, which has been operating for 15 years and is most closely tied to YouTube but has branched out to work with a variety of streamers and digital video players, including Amazon Fire and Roku, is moving past what’s been a largely manual and therefore labor-intensive and time-consuming process in the past by applying ChatGPT to its brand safety and suitability work. 

David George, Pixability’s CEO, explained that the use of AI enables far more accuracy and scale to predict and identify safe and suitable content environments for brands on YouTube and elsewhere. But it also helps to speed up the process dramatically. He declined to say how much the company has invested in the tech.

“This key investment has always been about dealing with the massive amounts of data and video that we harvest from Youtube every day, and that’s only growing” said George. “A lot of it is just being able to continue with the scale of what we do and do it in as timely a fashion as possible. We’re now supporting clients around the globe.”

Jackie Paulino, Pixability’s chief product officer, added the company has been using ChatGPT for about six months. “We’ve been applying this and using it to improve our training data for machine learning models,” said Paulino. “For YouTube particularly, it’s about speed to getting things into market. There’s lots of new trends that pop up on YouTube. So very quickly, being able to identify that something is new, and then being able to train a model on it, and then deploy that to find that those videos at scale, is really important.”

As an example, Paulino described setting brand safety standards around arms and ammunition — certainly a hot topic across all social and video platforms, but also one many advertisers want to avoid. “ChatGPT understands every gun that has ever been invented or used, so it’s much better than humans at categorizing that type of content,” she said. “Even for very niche things, a human may have mislabeled that in the past. Now, we know we can get that level of accuracy that’s actually better than humans can do.”

John Montgomery, an independent brand safety specialist who for many years worked with GroupM, said it’s the speed that really helps, because so much more content can be assessed in far less time. Generative AI “can go through such a huge volume so quickly, that it’s much more thorough and much, much more accurate,” he said. 

But Montgomery noted that human involvement needs to be maintained at the beginning and the end of any work generative AI does. “It’s only as good as the information you feed it, and the more detailed the brief that you give it, the better,” he said. “I don’t think that this is necessarily going to take human brain safety experts or brand safety reviewers out of the equation.”

Has this reached the agency world yet? Not quite, admits Jason Hartley, head of search, social and shopping at independent media agency PMG. “There is no doubt brands are cautiously excited for the incredible potential of generative AI, but as with any new technology, making the conceptual operational will take work and time,” said Hartley. “As this gets underway, it will be important for brands and agencies to work on the doable while preparing for the possible.”