I would actually had missed this whole observation and hence post if a collegue of mine would not had made a good post that we often look too far in the future instead of focusing on the here and now. He suggested we should start create the Smart Society for example by just making our office Smarter as a start. This whole thought of his sparkled my mind while out walking and realizing the album Spotify had just proposed was the perfect soundtrack for when the Digital and Physical converges.

The album is with my musical taste the perfect blend of acoustic and electronic instruments and hence is the musical representation and soundtrack for the convergence

The whole thing also made me realize that the Spotify despite my mix of many allas narrow musical preferences very driven by my listening mood nowadays with four very different Daily Mix most of the time makes a good jobb. That also includes in most of the mixes to avoid overly lyrically driven songs even if I do not only listen to instrumental material as well as my inclination to discover new music. Still a couple of times per month it will throw me something I really do not like but it is definitely worth the hassle for all the good new music it helps me discover.

Spotify Algoritms and I go way back

Up until two years ago the Spotify algoritms and myself where not the best of palls. My music taste as described above made any deep learning algoritm fail and would mix song that for me fits to different moods. I know my friends over at Spotify has made a lot of effort the adress that and for me it know works most of the time.

On the other hand the progress is not only to the merit of the Spotify team. I have since the start of their radio stations I think back in 2012 spent many hours or swiping away songs I did not enjoy and I would expect there is a waste library of what I have liked and not to data mine.

AI learning from this story?

Well actually more than it could appear so let me summarize it in 3 observations

  1. Developing good AI requires time and effort in a persistent continuous improvement mindset. Just consider the time, resources and computing power for the Alfa Go team to make a AI play a boardgame.
  2. For the final tuning you need human interaction to help the AI when it has not done a good jobb. Just like in my Spotify story where even if the AI has analyzed thousand of my listening hours from many years of consumption of the service it still needs intervention from my side to guide it.
  3. AI will in the short future be consumed by us a individual and companies through existing platform providers. They are the one to have the user scenarios where to apply AI, they have the user base to justify the investment and they have the large amount of data needed to feed the AI with

Sadly I think many companies are now engaging into the AI domain not being clear on the above observations and hence are quite likely to fail. But hey, we could assume that is just part of the AI hype cycle

Curious on the soundtrack that this is all about? It was Portico Quartet and the album Age of Automation

Join the Conversation

1 Comment

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.