In November 2015 Google released TensorFlow, a framework that can be used to create machine learning models and programs. The impact was more than tangible: it was explosive.
Before TensorFlow, the skills barrier required to specialise in machine learning was incredibly high, restricting usage of machine learning techniques to big tech companies and well funded government agencies. You had to be a real specialist, with a very specific academic background, before machine learning could become the focus of your career.
With the launch of a development framework, however, a lot of the more complex hurdles involved with machine learning were overcome. To be clear, TensorFlow hasn’t made the creation or use of machine learning a simple task; you still need to be a highly competent developer to create applications, algorithms or analysis tools that rely on the technology. But it has reduced the barrier to entry, making it cheaper to continually utilise machine learning whilst reducing the upfront cost of getting involved in the first place.
It managed to make tricky, highly technical concepts both accessible and fast to employ. Development time dropped from months to days for complex products. Much like the success Docker found with making encapsulation mainstream, what TensorFlow achieved with machine learning enabled the cost of experimentation with related techniques, tools and models to plummet. No longer tied to the need for a huge amount of technological infrastructure, or a terrifyingly deep wallet, machine learning became attainable and began attracting a much wider audience.
The result, as mentioned above, was explosive. New tools, techniques and applications began appearing; machine learning became a hallmark of innovation, with start-up after start-up forming around the technology. Along with that growth in investment and application came a job drive, with a rapid increase in recruitment for Machine Learning Developers, Machine Learning Engineers, Machine Learning Specialists and a host of similar roles, each previously niche yet now fairly mainstream.
Market trends for job titles associated with machine learning roles. Note the change in expected trajectory towards the back end of 2015, neatly tying in with the launch of Tensorflow? [*]
But it wasn’t just machine learning that benefitted; TensorFlow had a huge impact on the wider Artificial Intelligence community as well. Now, this is admittedly a little of a chicken and egg moment: it is plausible that a background increase in AI research was occurring, so TensorFlow became a viable product and was therefore created. There’s definitely some truth in that, but still, AI job searches, hires and skill sets see a similar spike in demand around the same time as machine learning roles, so there’s clearly some level of correlation.
As with the previous graph, this shows market trends associated with AI job roles; again there is a significant uptick around the back end of 2015.[*]
Plus, the real world impact cannot be downplayed. Machine learning has become a mainstay in the 24-hour news cycle, with new stories seemingly breaking every day. Associated techniques have been used in everything from competing in video game tournaments to making healthcare recommendations to influencing major political shifts. Machine learning, something that less than a decade ago was largely relegated to academic papers and science fiction, is now everywhere; it’s mainstream, common place, normal.
What happened with TensorFlow may be about to occur again. In July 2018, Google announced Cirq, a whole new open source framework with a very different target. Instead of AI or machine learning, Google has focused on another rapidly advancing technology that is set to fundamentally change the way we live: quantum computing!
The why behind Google’s ambitions is quite simple. With TensorFlow, Google created the go-to system for creating and maintaining machine-learning-backed software; in and of itself, that generated very little by way of revenue, but it gave them a captive market. By 2017, thanks to some clever integrations with their Cloud platforms, Google had been able to monetise that market. Sure, part of the appeal of TensorFlow is that it’s fairly device and service agnostic, but if it’s easiest to implement on Google’s services then that’s what a lot of people will choose.
On top of which, TensorFlow hasn’t ever been surpassed. Whilst there are some alternatives, most lack a lot of the functionality that Google’s framework provides and none come close to the market share that it has built. Surveys continue to place TensorFlow in the top two requested skills for machine learning roles, and it frequently ranks in the the top ten most desirable job experience traits for artificial intelligence positions*.
Right now, quantum computing is rapidly advancing to a comparable position to that which machine learning and AI had attained in 2015. By releasing Cirq ahead of that curve, Google are way ahead of the game. We can’t know what that will ultimately mean, but if TensorFlow is anything to go by we could be in for an interesting couple of years.
Eight years ago “Product” jobs were largely non-existent, and even where they had begun to appear their role and responsibilities were often fluid and non-uniform. Flash forward to 2018 and Product is now a major job-sector, helping to facilitate an industry-wide transition towards Agile working methodologies and resulting in the slow loss of more ‘traditional’ roles, such as Project Management. Most people have now heard of a Product Owner and Scrum Master, with their application even beginning to branch out of the tech sector, prompting write-ups on how the Product structure works and why it exists.
There are similar parallels in the more immature DevOps and Data Science fields, both currently entering a phase of formalisation and becoming increasingly mainstream. These new disciplines have sparked their own miniature revolutions, taking over the roles of database managers and system administrators, both of which are now on the decline.
Even the rise of Big Data has had a major impact on the job market, as the traditional role of Business Intelligence becomes squeezed by the need to react to (and the availability of) giant public datasets rather than carefully curated internal ones.
What each of these examples have in common is simple: a change in technology or technical thinking. Each demonstrate the impact that has historically occurred within the job market as a result of, comparatively, relatively minor shifts in resource availability or changing social attitudes.
When placed in comparison, quantum computing could be much, much bigger.
The concept of quantum processors is just a little terrifying. If faster, more powerful infrastructure has lead to both the Big Data revolution, and the more recent machine learning and AI revolution we’re currently in the midst of, then switching from bits to qubits will be huge. With a base-4 system in play, computational power increases on an exponential scale; with as ‘few’ as 300 qubits a quantum computer can perform more calculations that there are atoms in the universe.
That level of processing power will likely lead to a rapid expansion of our current techniques, enabling much faster data analysis, higher permutation deep learning algorithms and vastly superior modelling. The trick will be in getting the skills and tools in place to utilise these new resources, bringing that barrier of entry down as rapidly as possible – the same trick that TensorFlow managed with binary-based machine learning.
Right now, quantum computing jobs are few and far between, mainly focusing on research positions at learning institutes and R&D departments at the likes of IBM, Apple and Microsoft. Most still require a PhD, or equivalent, in quantum mechanics, quantum information science or similarly restrictive subjects.
However, as Cirq matures and the cost of quantum hardware decreases, the opportunity and conditions for another TensorFlow-like explosion are present. For now, all we can really do is sit back and watch, but the next wave of cutting-edge job campaigns could almost be upon us. As we’ve seen, once a new technology becomes accessible it causes big shifts in the job market and can change the nature of tech employment; whether that happens with quantum computing is not up for debate, only how quickly it occurs.
*All statistics and survey result data, including both graphs, have been pulled from ITJobsWatch under a Creative Commons license and are relevant as of 05/09/2018.
Hero image designed with the Vecteezy Editor