Hip Hip Hurr-AI at Infocomm 2024

PK on AI at infocomm

I’m sure that no one is surprised that artificial intelligence (AI) is the buzzword at Infocomm 2024, but is it secure? And does it work? Let’s dig in.

The first session I attended Wednesday at Infocomm 2024 was called “Chat GPT for AV.” The presenter showed how the platform could be used to answer a typical request for proposals (RFP) by uploading the RFP document directly into ChatGPT. I was impressed at the quick response, but it also seemed rather generic; aren’t sales proposals supposed to help differentiate one company from another? If all the bidders used ChatGPT to generate a response to the RFP, won’t the proposals all sound the same? Also, how would the project owner or consultant feel about uploading the project documents into Chat GPT? Aren’t those proprietary documents?

There was no hesitation about proprietary information on the part of the presenter, I guess they felt that once an RFP was out, anyone can do anything with it. Even if there is a disclaimer on the documents, nothing is stopping someone from uploading the documents to ChatGPT. Taking this a step further, nothing is stopping a hacker from pretending to be a qualified AV integrator. If the RFP response is simply generated from the RFP, it will answer all of the RFP questions in the most positive light, meaning it may not reflect the actual skills or qualifications or experience of the bidder. If the RFP says the company needs to be in business for at least five years, then ChatGPT will say it has been in business for over five years. This will likely lead to false qualifications. These have been a problem before but now with AI, it’s much easier to lie about who you are, or what you are qualified to do.

The presenter also demonstrated that ChatGPT is not very good at telling jokes, because as I always say, robots are not funny. He went on to show how ChatGPT could write something in the style of Dr. Suess. But did Dr. Suess give them permission to write about random subjects? Isn’t this the same as Scarlett Johassan’s voice being used without permission? And more importantly, is the Chat GPT generated poem anywhere near as good as one of Dr. Suess’ famous books, such as “Oh The Places You’ll Go” or “And to Think That I Saw It on Mulberry Street?” Of course not! It’s just a waste of time and computer resources, in my opinion. I had to dip out of the session early, but there are other AI sessions I plan to hit up before the show is over.

I mentioned Xten-AV in my previous column. Like many other booths, Xten-AV was talking up its AI capabilities. The company explained how AI is automating a lot of documents, and helping to refine its product library. Xten-AV’s AI model understands what you are doing and can suggest products that might fit in your project. A good example could be an adapter cable that is typically ordered with a piece of gear. XTEN-AV also uses AI to generate preliminary AV equipment rack layouts.

The company’s use of AI is not actually something new, though; XTEN-AV has been utilizing this sort of AI for about five years. It’s just being advertised as AI now because AI is so trendy. I really liked how XTEN-AV had solid examples of how AI can help AV folks do their jobs faster, not just write up generic documents. There were no Dr. Suess imitations or hallucinations. Also, Xten-AV’s products are fully implemented, tested and ready to go, unlike some of the AIvaporware I am seeing on the show floor.

See related  The Rainbow AI

image 67194369


One of the newer AI-enabled technologies that I learned more about this week is from Merlyn Mind, which is being showcased in the Newline and Samsung booths. Merlyn Mind is basically AI-enabled remote control for education; it can work with teachers’ laptops and other technology in the classrooms. It works a lot like the Xfinity cable TV remote: you push a button on a handheld remote and ask a question, then release the button. What sets Merlyn Mind apart from other solutions is its security features, allowing teachers to search online content without fear of hitting any adult websites or other not-safe-for-school pages. Merlyn can read a presentation that is on the screen and suggest similar content, and its large language model is pulling from trusted educational content, not the entire Internet.

According to the Merlyn Mind webpage, it does not sell personal information, retain voice audio after processing or use voice audio to identify individuals. This voice recognition feature is something other AV solutions are flaunting without realizing the privacy, security and equity concerns that voice identification can raise. This sort of emphasis on security and privacy makes me happy.

So, how well does Merlyn Mind work? Let’s just say a trade show floor is not the same as a typical classroom.

image 50393089

When I tried the Merlyn Mind remote control in the Newline booth, it had trouble separating my voice from the other voices and noise on the floor, even when I tried slowly shouting. This is a common problem with any new voice control. The microphones in the handheld remote are often chosen to be small and low cost, not high-fidelity, and the system relies on AI and digital signal processing algorithms to separate the voice from the background noise, and translate the voice into text. Add in a constant barrage of background noise from a tradeshow like Infocomm, and the demo equipment that might have worked flawlessly in the test lab doesn’t perform very well. Background noise aside, the Newline Merlyn Mind demo did work and, overall, I was impressed.

The Samsung booth experience was a little different, and the demonstrators admitted the new interactive whiteboard product was a prototype. Samsung says, “New AI-powered features include automatic transcriptions from spoken lessons and the ability to generate detailed class summaries that highlight key points and main topics. By analyzing the teacher’s voice transcription, the written material on the display and provided educational materials, it can also generate quizzes for students. It also uses machine learning to improve content accuracy and block inappropriate content. Additionally, through a collaboration with educational technology company Merlyn Mind, the display will feature voice recognition technology optimized for educational environments. This will simplify control of the Interactive Display and help tailor its functionalities to align with school-specific curricula.”

Keyword: will.

Much like the Merlyn Mind remote had trouble figuring out what was being said in the Newline booth, the remote in the Samsung struggled to understand my voice even when speaking slowly and loudly. Moreover, the Samsung demo system could only search what was on the screen, it was not able to search for any new topics. It had been set up only for careful demonstrations using content about amphibians. While this is a common practice at Infocomm and other trade shows, it took away the wow factor. It was sort of like shopping for a TV, and the salesperson telling you that only one channel was available.