The Best Song in the World

Written by Paul Siluch
July 31st, 2024

What is the most creative song ever written? The most finely crafted?

And why am I asking this question?

Music will always be judged in the context of when you lived. For example, Beethoven superseded Bach and Handel in classical music, as did Mozart and Debussy after him.

In modern times, the Everly Brothers’ “All I Have to Do Is Dream” is recognized as one of the most perfect songs of its era, as were the Beatle’s “Yesterday” and “Strawberry Fields” a decade later. Aretha Franklin’s “Respect” and Led Zeppelin’s “Stairway to Heaven” broke new ground of their own the decade after that.

We like what we grow up with.

Artificial Intelligence is now being used to write music – the newest realm to be conquered by machines as AI changes the world faster than we can comprehend it. Companies offering AI solutions to the world through chatbots and new algorithms have soared in value as investors rush to bet on these new winners.

The Belmont Stakes is the third race in horse racing’s Triple Crown.

It is 1 ½ miles long. Ask anyone who has raced this distance – including any horse you know – and they will tell you it is a long distance to run flat out. It is still early in this race to Artificial Intelligence. The horses leading today may not be the eventual winners.

Party Like It’s 1999

In 1999, a hedge fund manager named Nandu visited our office. It was near the peak of the dotcom boom and every stock involved with “the internet” was flying higher. Especially the fibre optic cable stocks. Cable-laying ships crisscrossed the oceans to lay glass wire everywhere. We even had a ship with a giant spool dock in Victoria for a few weeks.

Nandu and his firm had calculated that companies had wrapped the Earth four times with cable, which was impressive, but there was one problem: the cables were mostly empty. It was as if we built four new TMX oil pipelines and only had enough oil to fill half of one.

The cables were empty because most of the internet traffic was simple text for e-mails, faxes, or long-distance phone calls. Not dense, in terms of data volume. With so little data and so much spare capacity, companies couldn’t charge more to pay for all the cable. Pretty soon, sales for new fibre slammed to a halt. Telecom companies who ordered extra cable suddenly dumped theirs on the market and a glut developed.

The rest, of course, is history. Nortel, JDS Uniphase, Lucent Technologies, and many others either went bankrupt or collapsed in price.

The cables were empty, but they didn’t stay empty. Google went public in 2004, ushering in an era of something called “search”. Facebook went public in 2007 and started the social media craze, and YouTube gobbled data with millions (now billions) of videos. By about 2010, all those fibre-optic cables were full and more needed to be laid. It was too late for Nortel and the early pioneers, but the cheap data pipes were a blessing for the new entrants.

The Internet Boom began with a hardware “pick and shovel” phase. But, like the gold rush of 1849, few of the original miners who got rich stayed rich. It was software companies that used the internet that got rich and stayed rich.

Today, we have a similar era with Artificial Intelligence. Almost every industry (or so it seems) is adding AI to its products and services.

• Tech support? AI to the rescue with our helpful chatbot.
• Improve your writing? AI is your new tutor.
• Need an illustration? Craiyon, Illustroke, OpenArt and a hundred other graphic AI companies are ready to draw anything for you.

Companies are ordering millions of the most expensive chips – some the size of a paperback novel weighing almost 4 pounds - costing over US $100,000 each. Everyone is racing to be the best at AI, whatever it means.

But even as everyone adds features like conversational search and smarter answers, companies are struggling to find ways to charge for them. For example, a Google search is essentially free. A ChatGPT (or Gemini, or any other AI) search costs between $0.01 and $0.06. It costs Microsoft close to $500 million per year (source: Noah Johnson) which they do not get back yet.

Some people will pay more to use AI, but to be honest, free Google is still pretty good. I can get by. So can most people. For now.

At some point, companies are going to realize they can’t keep ordering $100,000 chips with no offsetting revenue. And then they will stop, bringing the chip titans back to reality. When? Another year or even two? Every boom is followed by a bust. It happened with railroads, radios, gas stations, and coffee shops. We get carried away and build too many, and then we stop.

Gartner Group estimates that 30% of Generative AI projects will be abandoned by the end of 2025.

“Never make predictions, especially about the future.”
- Casey Stengel

The AI boom is real and cannot be ignored, for we have no idea what it will eventually become.

Who could envision SpaceX when the Wright Brothers first flew at Kittyhawk, for example - but it is still very early days. It took years for commercial aviation to become a reality.

Back to the music question. Can AI write the next big hit? Be the next Beatles?

Ever hear of the bands Klaatu, Big Star, or the Flaming Lips? I didn’t think so.

These were bands that sounded like the Beatles. Generative AI systems predict what should come next based on past data. If you feed the model only Beatle’s songs, it will create similar ones that sound the same but different. Like the bands that tried to mimic the Fab Four, Generative AI copies and recreates what is already there.

True creativity comes from experimentation, emotional pain, and the Eureka moment that AI is not yet close to. AI can help song writers by automating certain tasks, but human creativity and innovation cannot yet be replaced. AI will become a productivity tool like Google Search or Excel, helping humans to be smarter and faster.

For the record, Rolling Stone’s 2024 survey of the Greatest Songs of All Time listed Bob Dylan’s 1965 “Like a Rolling Stone” as #1.

Hurry Up and Wait

Some problems will take years to figure out, even with AI. Like autonomous driving. How long has that been promised?

The reality is that machine learning is here, it is accelerating, and it is being widely adopted in every industry. John Lennon wrote his songs using new technology - an electric guitar and digital sound equipment – after all. But it was still John who wrote them.

Why won’t humans be replaced yet? There are still many problems to overcome. Some intelligent systems have ‘hallucinations’ and offer crazy answers. Some have even been caught lying!

Machines trained on information they themselves produce gradually go insane in a process scientists call ‘data inbreeding’. Yes, they produce amazing pictures and remarkable essays, but they do not yet truly create.

Meanwhile, look at what we neglect. Calgary has 50-year-old water pipes collapsing that should have been repaired or replaced years ago. We need so much more electricity for all our heat pumps and electric bikes but can’t open enough copper mines or make enough transformers to get the power to your house. Basic stuff has been neglected. The turn in the markets from the tech giants to the small cap industrial stocks may be reflecting this.

That’s the next boom, but it may have to wait for this one to end first.