Shiny Objects

keeping up with the joneses - software style

The Silver Bullet fallacy (or magic bullet syndrome) refers to the odd tendency of people to seek a single, simple solution to complex problems, often believing that this solution will solve all their issues effortlessly. In reality, these solutions are as rare as my pet unicorn (which incidentally, requires much more upkeep and effort than you’d expect).

Didn’t Work, Did It

I spent a chunk of my life playing and making music. Time and time again, I’ve watched mediocre musicians buy hugely expensive instruments, expecting that a $5000 instrument would magically make them better, only to struggle even more with a professional instrument. In fact, I’ve recently been playing my Buffet R13 clarinet fairly regularly, and it’s been a bit of a struggle - partially because I’m rusty, but partially because pro instruments require better technique, and I’m rusty.

Also from the music world are the folks who pay hundreds of dollars for a master class when a few pointers from a $50 lesson with local teacher would do much more to help them grow from where they are now.

There are no shortcuts.

Money as a Problem Solver

I was listening to the Pivot podcast this morning, and both hosts were talking about how they (both are extremely affluent) don’t want to send their children to expensive private schools, because they don’t necessarily offer a better education, and that investing that money, instead of paying for school is better in every way in the long term. Admittedly, private schools exist because of inadequacies in US public education, but enrichment from activities outside of school (sports, play, social groups) are often undervalued.

It reminds me of a huge takeaway I took from David and Goliath, by Malcom Gladwell. For a few chapters in the book, Gladwell talks about the competitiveness of prestigious universities and shares stories of extremely intelligent and talented students who drop out or fail - because as good as they are, it’s too competitive for them to succeed. Eventually, he reaches a conclusion that dispels the magic of what universities offer - rather than go to the absolute best university one can get into (or afford) - and being merely “good” or worse there, you’re much better off in the long run going to the best school where you can be a top student.

That the best students from mediocre schools were almost always a better bet than good students from the very best schools.

Gladwell discusses the psychology behind this phenomenom, and admittedly, it isn’t always about throwing money at chance to succeed, but as someone on the board of a liberal arts college, these stories hit the same to me.

The Magic of AI

The third of Arthur Clarke’s Three Laws is:

Any sufficiently advanced technology is indistinguishable from magic.

People want magic to solve their problems, and the current hotness in advanced technology (and magic) is Generative AI. For far too many people, AI is either the solution to every problem they can imagine - or the thing that will take their job away.

It’s neither.

Generative AI is as much an algorithm as it is AI. Technically it creates new content based on what the algorithms have consumed from massive amounts of data. But it’s not magic. It’s a very efficient parrot that can take what it has “learned” and present it back to you - flaws and all.

History Repeats

Sometime in 2013-2015, I gave a talk at an internal Microsoft conference about Data Science. Data Science was the magic of its time, and everyone wanted to know what to do with it, and how they could “give all their data to data scientists” in order to get their unknonwn problems solved. I remember pulling search trends from google trends at the time - while I no longer can find that presentation in my archives, I find it interesting to see how much interest has grown since then.

The Rise of Data Science

What do you call a statistician who lives in Silicon Valley?

A Data Scientist

Unattributed joke from 2013

At the time, Data Science was new, and while it felt like magic, we eventually learned that it wasn’t. In fact, when I talked with a few of our Data Scientists in the XBox org, I discovered that they spent the vast majority of their time cleaning up data, and very little “sciencing”.

As expected with the previous round of magic, Generative AI is no different. While GenAI isn’t impacted by the form of the data as much as Data Science, it’s highly impacted by the quality of the data. It’s just not smart enough to throw out misinformation, or incorrect “facts” from the data sets it consumes.

For complete transparency, it’s worth stating that I’m a huge user of ChatGPT and find it invaluable for many things I do (and continually discover more uses for it). It’s a helpful brainstorming companion (and today, helped me figure out that the Gladwell book I was thinking of was David and Goliath, and not Outliers!).

It’s an extremely helpful tool, but not magic.

No Silver Bullet

Almost forty years ago, Fred Brooks wrote No Silver Bullet - Essence and Accident in Software Engineering. In that essay he calls out that many people in the software industry are in frequent search of “silver bullets”, such as new programming languages, development methodologies, or tools, in the hope that they will dramatically improve productivity or eliminate difficulties. However, he asserts that these supposed solutions often fail to deliver the promised benefits because they oversimplify the complexities of software development.

We may be doomed to repeat our own history far beyond my career-span, but in the meantime, hold on tight, and don’t forget to think.

-A 1:1

P.S. - my friend ChatGPT told me the following - which is probably worth sharing as a disclaimer.

It's worth noting that some statements are subjective or based on personal experiences/opinions rather than objective facts. For instance, the effectiveness of private schools versus public schools may vary depending on various factors and perspectives. Additionally, while the author discusses the limitations and challenges of Generative AI and Data Science, their characterization of these technologies is generally accurate, although some nuances might be oversimplified.

Yep - sounds about right.