News

Soundbyte 333: Moonlight Sonata

18 February 2018

Another tumultuous week in the world of politics. One Dutch member of cabinet had to resign due to lying through his teeth, but that’s not even the biggest event. Ever since FBI director Comey was fired May last year, I developed an unhealthy obsession following the news around Robert Mueller’s investigation into election interference in the US. Last Friday a bombshell story dropped: indictments of several shady Russian firms and persons who actively engaged in manipulation of another sovereign nation. Not the first result of the investigation, but definitely the most shocking one so far (although I’m confident much more will follow). Earlier, we also found out our own Dutch intelligence service played a role in uncovering parts of these actions.

On the whole, nation states manipulating the populace of other nations is of course nothing new. What is new, however, is the sheer magnitude and impact of these online operations. And it doesn’t look good for our favorite tech platforms. Facebook, Twitter and undoubtedly other social media were essentially weaponized. Identity theft and coordinated disinformation campaigns raged on a scale never (openly) seen before. And by the looks of it, it worked. A significant amount of people happily consumed bot-fabricated information, aligning with their pre-conceived notions. The filter bubble is real and is being actively abused.

This got me thinking. Facebook, Twitter, and others, all use advanced machine learning models to steer (dis)information towards users. The algorithmic news feed, suggestions who to follow, everything is decided by algorithms. Problem is, these algorithms worked ‘as designed’ yet they led to very undesirable results, to say the least.

When I was asked by InfoQ what the most important skill for software developers to learn is for the next years, I immediately answered data science/machine learning. But, the above usage was not what I had in mind. Obviously, any technology can be used for good or bad. Unfortunately, this time the creators of the technology were not the ones pulling the trigger (I hope). Tech got played. Still, ignorance is no excuse.

On a more optimistic note, I do still believe the good of machine learning will outweigh the bad. That’s why several of us at Luminis Apeldoorn are doubling down on learning this technology, for example by following the Coursera Deep Learning specialization. In five courses of several weeks each, the theory and application of deep learning models is taught. At the other end of the spectrum we’re also running a ‘Machine Learning for Dummies’ session for non-technical Luminis employees next month (let me know if you want to join and haven’t registered already). We already see many great applications of ML in current projects, with more to come. An exciting field indeed, but as we’ve seen, not without (moral) hazards.

Let’s end with some music. This time an amazing performance by a teen prodigy, combining Beethoven and electric guitar. Enjoy!

One Response to Soundbyte 333: Moonlight Sonata

  1. Wim Offermans says:

    I’am happy to see that you at Luminis are turning your attention to what I see as a main threat to our freedom.

Leave a Reply

Your email address will not be published. Required fields are marked *