As AI transforms various aspects of content creation and distribution, companies are also having to deal with challenges such as AI hallucinations that are originating from limited training data and data biases.
Last year, the Cambridge Dictionary declared ‘hallucinate’ the word of the year owing to the new connotations it had taken on with the rapid uptake of AI across different sectors. Al has become an integral part of our corporate life. The media and entertainment industry has particularly employed AI for subtitling and instant transcriptions, with face and audio recognition cutting down the time used for this drastically while also being cost-effective.
But on occasion, users have encountered false or misleading information generated by AI, and these mistakes crop up more often in regions like the Middle East because the AI models we have at present are all trained on datasets largely dependent on Western sources and the English language. As a result, several biases have crept into the AI models. Termed AI hallucinations, these errors have become a challenge for media companies in the Middle East and Frederic Petitpont, co-founder and CTO of Moments Lab, formerly Newsbridge, has addressed this quite aptly in this month’s issue of BroadcastPro ME.
With AI reshaping our world all the way from content creation to delivery, it is critical that we understand it more in-depth. And what better place to discover this than at NAB Show, which is just around the corner. There will be extensive discussions on the tech advancements, opportunities and challenges around AI at NAB. You may also want to attend the new Main Stage discussion on AI in Media and Entertainment at the show, which will bring together some high-profile market majors to discuss their approaches to AI, including the merits and pitfalls of its ongoing and expanded use in production and development. There will be a whopping 120 sessions on AI at this year’s show from April 13-17. See you in Vegas.
