One cool and one not so cool use of AI in IFE.
Image courtesy of Chat GPT
I was initially going to make this month’s post about how we are currently testing some very cool AI tech to upscale old tv and movie content to 4k for our customers but then I did some googling on AI in inflight entertainment. What I found was a mix of interesting ideas and regurgitated GPT based blog posts.
Not So Cool.
When most people think about AI today they are thinking about Chat GPT or Google Gemini. These are large language models that are trained to understand human like conversational writing and then be able to answer in kind. With their answers they will take their best guess at giving you the answer you want to hear.
I came across a blogpost over on this aviation blog: here. It is titled
‘The Role of AI in Modern In-Flight Entertainment: A New Era In Personalized Travel’.
I hope I am wrong but the content and structure of the piece stinks of GPT. Don’t get me wrong, I am no writer and I will happily use AI to lay out a structure for me but these kind of articles, if written by AI, only serve to try and boost your websites SEO. They aren’t adding anything or creating dialog. As an experiment I have asked Chat GPT to write an article on the same title here. You can compare them.
One suggestion for the future of IFE the blog suggests:
AI can adjust content based on real-time passenger behavior and feedback. For instance, if a passenger frequently pauses or skips certain types of content, the AI system can learn from this behavior and adjust future recommendations accordingly. This dynamic content adjustment ensures that passengers always have access to relevant and engaging entertainment options.
Thrilling.
While this isn’t a direct use of AI in IFE it does speak to a trend of buzzwords I have experienced in the industry for the last 15 years. Before it was AI is was ML. Before that, big data and before that Web 2.0. These often lead to empty features like a skinned version of Chat GTP as a customer service bot.
Image of the Ryff platform from Ryff’s website.
Cool.
More interesting to me was this article from Anuvu here. TLDR for those who haven’t read it is that Anuvu are partnering with Ryff to provide a bespoke extra layer to their QA and content screening level. AI will proactively recommend areas to review when the compliance teams are going through their process.
From what I can gather from Ryff’s website (see their summary of their platform) they use a specially trained large language model to categorize objects, people and context within video media.
At Jet Pack IFE we also have a rigorous QA process - we operate in many territories around the world where there are different attitudes to what is acceptable media to show people. I can, and I think my team would agree, totally appreciate where an additional step to proactively suggest things we may have missed could be valuable. Ultimately the passenger ends up benefiting as the chance of them being exposed to inappropriate content or perhaps bad metadata translation is reduced.
Whether it is checking video for inappropriate content, translating metadata or crunching usage statistic numbers it is very important for there always to be a layer of human oversight because as I mentioned earlier and as Kosh mentions in his article, large language models will only ever take a guess, even if it is a very well informed one.