5 examples how 2019 showed us that exponential technology is more than just a catchphrase

As a technology optimist and engineering physicist, I was always very sceptical about catchphrases like “the pace of technological change is becoming exponential” or “the impact of digital disruption”. They often sound pretty meaningless because they’re mostly used to sell you expensive consulting hours. However, in the past two years, I really began to see the first impact of technological change at a pace that, for me, became harder and harder to grasp.

McKinsey research states that today, 47% of all current tasks can already be automated with technology that is currently available on the market — that’s almost HALF! And we’re not only talking about the repetitive tasks of factory workers or administrative jobs… this is happening across all industries and skill levels as we’ll see now.
Let’s have a look at some (cool) AI examples
OpenAI, the research institute for Artificial intelligence (co-founded by Elon Musk), recently published a new AI language model that has learned to generate text from scratch and excels at text comprehension, machine translation, question answering, and summarisation.
The whole purpose of openAI is to open source all their research because they believe that by commoditising AI they are increasing the chance of a good outcome of general AI. This was the first time they didn’t release an open source version of the trained model, because they were afraid of its impact (fake news, etc…), which demonstrates the power these AI models are beginning to show.
Still, openAI open sourced the supporting paper and the underlying code, and shortly after, another research team re-trained a copy of the GPT-2 model and now this technology is available for anyone to use. If you want to experiment with a weaker version of the model (less parameters) that openAI did open source, here is the link: https://talktotransformer.com
Being a machine learning engineer myself, I have to admit that I was shocked by these results. Text generation used to be a very challenging task and most of the researchers thought it would take a longer time to crack the code. To show the speed at which this progress was made, below is a sample which was generated by a state of the art model (trained on the Harry Potter series) in 2017:
They had no choice but the most recent univerbeen fairly uncomfortable and dangerous as ever. As long as he dived experience that it was not uncertain that even Harry had taken in black tail as the train roared and was thin, but Harry, Ron, and Hermione, at the fact that he was in complete disarraying the rest of the class holding him, he should have been able to prove them.
In only two years time they went from this incoherent gibberish to high quality, auto-generated text, on multiple domain-specific topics. I don’t think I have to spell out the impact that automatic text generation will have on the fields of copywriting, content marketing, journalism, interpreters…
But this reaches even further. As this research was open sourced, it didn’t take long before Google implemented similar algorithms in their search engine. The whole SEO world is now trying to figure out how it will impact their business. This example shows how a breakthrough in one research field, gets amplified and starts impacting other fields.
And this was only one example in one of the fields of AI research. Right now we’re also seeing more and more AI powered software that’s capable of combining huge amounts of data and automating end-to-end marketing, drip and CPC (cost per click) campaigns.
Apart from the marketing industry, also the user interface development profession is under “attack”. In this video we see how an AI model, using computer vision for object tracking, is transforming wireframes into fully functional HTML code, in real-time. “Design to code” as the creators like to call it.
And what about the creative sector? A close friend and ex-colleague of mine, Xander Steenbrugge, recently released this video, where an AI model is automatically generating images based on the input of the music. This model was built in less than 2 months, as a side project.
Nearly all of these examples are based on applying open-source software to different industries. This means that anyone with access to the internet and some technical background can build further on the progress made by others, which is drastically decreasing the time from research to production.
Before anyone starts feeling sorry for themselves because the AI research community is stealing their jobs, hold your horses! In the process of training these AI models, a lot of parameters need to be tuned in order to get a good performing model (model architecture, hyper parameters, training vs test set cut-offs, …). This used to be a job for experienced Data scientists and Machine learning engineers, but guess what? There are AI algorithms that can build and train other AI models better and faster, using a method called autoML.
The above examples are focussed on the fields of IT and marketing, but this is a widespread evolution that is also redefining healthcare, agriculture, education, biotech, etc…
What about the catchphrases “exponential change” and the “impact of digital transformation”?

When I realised that all of the above examples happened in 2019, I started to grasp the exponential curve at which technological change is happening. We are entering the second part of the hockey stick (a.k.a the exponential growth curve). People have been talking about this phenomenon for quite some time now. But it’s really in these past couple of years, that we are beginning to see the impact and pace, in real-time.
What’s next?
In my next article we’ll look at the impact of automation on the time a skill stays relevant, the so called “skill lifetime”. As you probably expected, we’ll see that the lifetime of skills is decreasing more rapidly than ever. But no need to despair, because we’ll also introduce you to a “game plan” that you can start applying in your career and business to stay on top!

