Making radio software suck less!
Results 1 to 6 of 6

Thread: Making radio software suck less!

  1. #1

    Making radio software suck less!

    https://www.linkedin.com/pulse/makin...tenburg-togaf/

    Working in the radio broadcast industry for years, it's noticeable software vendors are WAY behind the times in this industry. While obtaining advanced certifications in Artificial Intelligence/Machine Learning/Deep Learning programming I started to think, "Why doesn't traditional radio programming use this technology?". Brainstorming with friends and colleagues we decided to embark on the challenge of delivering a new and innovative software solution for the radio broadcast industry which would save money, time and deliver a superior experience for the advertisers and end users. Our first task is to develop a new music scheduler application.

    This cutting-edge application will revolutionize how radio is delivered and consumed. We will utilize artificial intelligence (A.I.) to pick songs with/without complicated "Rules". Customers will be able to use geographic, social media and radio monitoring data to enhance the A.I. decision. Programmers can share and contribute to other stations music rotation. This application will give you the option of running it in your web browser, Windows desktop, Mac, iOS, Android, and more. Development will be agile...meaning, once the MVP is ready, you will continue to see improvements daily.

    User experience is our primary focus!

    We are currently working on the MVP "Music Scheduling" software and would appreciate any feedback.

  2. #2
    Would love to get any feedback on this idea, is this something useful?

    Would you want an AI song selector or a combination of Rules and AI?

  3. #3

    Join Date
    Oct 2007
    Location
    Bellingham WA
    Posts
    6,040
    Quote Originally Posted by WhittenMike807 View Post
    Would love to get any feedback on this idea, is this something useful?

    Would you want an AI song selector or a combination of Rules and AI?
    Perhaps Rules and AI for now. Not all the bugs are worked out in AI to make it fully independent. You have to feed it some kind of formula. Then fine tune the bugs out as you go. Trouble is the sudden variables. For example; A song could be high on the charts nationally, but falling hard locally in favor of the artist's newest single climbing up the charts.

    The other is the viral songs: Good with a trained human eye on YouTube and Twitter who spots and gets them on the air as fast as they can. But with total AI, it may not be able to discern between a true viral pop song and a disgusting hate-filled political screed or scare mongering song pumped up by foreign or domestic social media bots to take advantage of the system if it's use is widely adopted.

    I think you're onto something. But in every new technology, there are overlooked or underestimated vulnerabilities. Especially if you have a social media end to the system, It's safest to be proactive and prepare for anything from a DDoS attack (especially if the app information goes through a central local web server.), to programmers gaslighting other programmers. Or bots doing that, angry former MDs, etc.

    If you're creating direct desktop apps, Linux is becoming widely adopted worldwide. The next Microsoft OS, Azure Sphere is based on Linux (Windows OS is being retired.) Make it compatible with at least Ubuntu 14.04 AMD 64 and above, keep a PPA repository (I know everything's going to Snap and Flatpak in 18.04, but they extended the security releases for 14.04 and 16.04 because they're still so widely in use and some older systems don't handle Snap and Flatpak very well. So PPA remains secure until at least 2022.)
    To the person who stole my antidepressants; I hope you're happy now.

  4. #4
    Quote Originally Posted by WhittenMike807 View Post
    https://www.linkedin.com/pulse/makin...tenburg-togaf/

    Working in the radio broadcast industry for years, it's noticeable software vendors are WAY behind the times in this industry. While obtaining advanced certifications in Artificial Intelligence/Machine Learning/Deep Learning programming I started to think, "Why doesn't traditional radio programming use this technology?". Brainstorming with friends and colleagues we decided to embark on the challenge of delivering a new and innovative software solution for the radio broadcast industry which would save money, time and deliver a superior experience for the advertisers and end users. Our first task is to develop a new music scheduler application.

    This cutting-edge application will revolutionize how radio is delivered and consumed. We will utilize artificial intelligence (A.I.) to pick songs with/without complicated "Rules". Customers will be able to use geographic, social media and radio monitoring data to enhance the A.I. decision. Programmers can share and contribute to other stations music rotation. This application will give you the option of running it in your web browser, Windows desktop, Mac, iOS, Android, and more. Development will be agile...meaning, once the MVP is ready, you will continue to see improvements daily.

    User experience is our primary focus!

    We are currently working on the MVP "Music Scheduling" software and would appreciate any feedback.
    I developed some music scheduling software back in the 80's which used a crude version of AI. It worked pretty well, but required being coupled to frequently-updated music research data. At the time, the only way you could get that data was from call-out research. As we've seen over the years, call-out-data from phone surveys is marginally accurate. Either one needs to pay survey takers to pay attention and give their accurate feelings about music, which would be cost-prohibitive, or some sort of per-song PPM data would need to be developed. Again, expensive.

    I can't remember the station, but it tried a form of AI using requests through their website. Just like the days of a dial-in request line, you had request-hogs that would pepper the station for (example) the latest Taylor Swift song. The whole thing flopped pretty quickly.

    Just like human intelligence, Artificial Intelligence relies on a lot of data input to develop behavioral patterns. Getting into the heads of listeners that could provide a broad enough spectrum of information to pick playlists from, is the secret sauce. Not the picking software.

  5. #5
    Hmmm, typically music scheduling software is something that's picked by ownership, not the rank & file. When they hire the programmer, they list in the qualifications that there is experience with their particular music scheduling software. So launching a new software, or even adaptions to existing software, probably would require the developer having a booth at the recently-completed NAB radio convention, and then making the rounds to national PDs at the major companies. Obviously iHeart has a vested interest in one brand, so that's probably a tough sell.

  6. #6
    Moderator/Co-Administrator
    Join Date
    Jul 2003
    Location
    SoCal
    Posts
    37,437
    Quote Originally Posted by WhittenMike807 View Post
    https://www.linkedin.com/pulse/makin...tenburg-togaf/

    Working in the radio broadcast industry for years, it's noticeable software vendors are WAY behind the times in this industry. While obtaining advanced certifications in Artificial Intelligence/Machine Learning/Deep Learning programming I started to think, "Why doesn't traditional radio programming use this technology?". Brainstorming with friends and colleagues we decided to embark on the challenge of delivering a new and innovative software solution for the radio broadcast industry which would save money, time and deliver a superior experience for the advertisers and end users. Our first task is to develop a new music scheduler application.

    This cutting-edge application will revolutionize how radio is delivered and consumed. We will utilize artificial intelligence (A.I.) to pick songs with/without complicated "Rules". Customers will be able to use geographic, social media and radio monitoring data to enhance the A.I. decision. Programmers can share and contribute to other stations music rotation. This application will give you the option of running it in your web browser, Windows desktop, Mac, iOS, Android, and more. Development will be agile...meaning, once the MVP is ready, you will continue to see improvements daily.

    User experience is our primary focus!

    We are currently working on the MVP "Music Scheduling" software and would appreciate any feedback.
    In a Radio World article last month by Dave Beasing, I was quoted as suggesting AI as a future step on music scheduling software.

    But my suggestion is not based on outside data, but, rather on making best song choices for segues and sets based on the history of log editing by the station programmer. In other words, the software brings up choices based on past patterns of optimum, edited scheduling.

    When we edit our logs (and, unfortunately, this is less prevalent today due to time pressures) we look for bad segues and massage flow to get the best-feeling music segments.

    The key is creating a feel for the station, just as the top DJs in the club world do. You don't see TiŽsto or David Guetta or Doplo using a computer to select songs and find ones that flow best together. Neither does a good PD.

    Pandora gives music set up by an algorithm. It does not fare well compared to the flow of a highly curated music radio stations. The disadvantages of traditional radio are not in music selection and flow, they are in high commercial loads and being on a platform that is increasingly unpopular.

    Music scheduling software is pretty much the palette that a good programmer uses to put together the colors they want to use. It does the complicated math of looking at vertical and horizontal rotations, same artist separation, etc. that are tedious to do manually. It lays out the "best songs" based on that kind of restraints, and then the PD can massage based on feel.

    We don't pick songs based purely on downloads. We don't look at the (often mistaken or geographically wrong) lists of other stations to program as their target and competitive array is different. We use our own personal evaluation of every song before it is added to make sure it is right for our station. We use the skills of a Picasso, not of a house painter, to pick songs and combine them on the air.

    There is current software that runs on mobile devices. I know of more than a couple of PDs who were not in their stations when Aretha Franklin passed to edited their logs from "the street" to honor her immediately.

    Back to the point I made in Radio World: AI has its best application in learning a programmer's style and emulating it when there are scheduling "best song" situations. But the system still needs rules to know the tempo of a song, its research score, its first airdate, the possible artist conflicts (Beatles / McCartney / Wings like issues), daypart restrictions, and all the rotational patterns and such.

    For decades there have been efforts to predict what songs are going to be hits based on AI. Were the concept to be workable, there would be no stiffs put out by the music industry. AI can be a tool, but not the engine for music scheduling.
    www.americanradiohistory.com
    Broadcasting Magazine and Yearbooks, Billboard, Cash Box, R&R, Record World, Music & Media, Audio, Television/Radio Age, R&R, Duncan's American Radio, Popular Electronics, Studio Sound, Broadcast Engineering, db, and more.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  




     
Our Conferences
Useful Contacts
Community


Contact Us