Training for the Artificially-Intelligent Future

The fastest way to leverage machine learning at scale may be to train internal resources in key data science concepts.

With many hot technology trends vying for attention, artificial intelligence and big data have emerged as two areas that seem certain to significantly impact how enterprises use tech. Currently, it seems like any company worth its salt is figuring out ways to leverage data science, machine learning and neural nets. And many have even started implementing this tech in the real world.

It is ironic that, when implementing tech focused on autonomous computing, the human factor can be the big challenge. Specifically, enterprises often struggle with gaining the expertise needed to make use of such complex and fast-evolving new technologies. While hiring staff or acquiring start-ups with existing expertise are tempting options, they are not always practical or even possible.

Increasingly then, companies are investing in getting their existing staff trained up for the artificially-intelligent future.

Study the Machine Learning Hype

Many employers have started offering training courses related to machine learning. And relevant online courses are among the most popular available, with new ones emerging regularly. This seems to suggest machine learning, AI and data science are increasingly vital areas for enterprises. But is this really the case and—if so—why is there so much emphasis on training, specifically?

It is commonly assumed that there is a huge amount of available data that could potentially be converted into business value via machine learning. Tech news sites are littered with stories on this subject, which often escalate from enthusiasm to hype to outright science fiction. So, is there really a huge supply of data that enterprises could theoretically leverage via machine learning algorithms?

The short answer is yes—many companies have benefitted greatly from this phenomenon and

some have built entire businesses around it. Machine learning has become so ubiquitous among start-ups that some no longer bother mentioning it in their pitches because its presence is a given. Being wary of new technology trend hypes is always advisable but there is something to this one.

To implement machine learning at scale, enterprises will need much greater numbers of staff with the relevant skills.

— Christopher Bonnell, Senior Principal Data Scientist, CA Technologies

Implementing Machine Learning at Scale

Given the success enterprises are experiencing with machine learning projects, it’s understandable that they would be extremely keen to scale up these activities as quickly as possible. But here is where that troublesome human element comes into play—to implement machine learning at scale, enterprises will need much greater numbers of staff with the relevant skills.

Why not just hire a bunch of experts? That’s not a bad question as there are certainly plenty of these experts out there right now. But the market for hiring data scientists above a certain skill level has become extremely fierce, with many competitors often fighting over a single applicant and turnaround times for jobs measured in days. So, hiring the numbers you need is going to be a challenge.

Data Science Experts vs. Machine Learning Practitioners

Lacking the required expertise is often a catch 22: If you lack expertise in a subject, it can be hard to tell a seasoned expert from a hack. So, not only will you have trouble getting the quantity of staff you need, you may also have trouble getting the quality (depending on your own level of expertise). This is made worse in the case of machine learning because of a key distinction among candidates.

Many data science experts aren’t really machine learning practitioners. Often, they come from academic backgrounds and focus on coming up with new algorithms to solve problems, rather than on actually solving specific, real-world problems. In reality, broad knowledge on the topic of data science may be less valuable than specific skills—the practical ability to actually do machine learning.

Hiring true machine learning practitioners is clearly a challenge. One way to circumvent this challenge might be outsourcing. In practice though, that route has proven to be a rocky road, fraught with expense and confusion. Essentially, we are left with one remaining option: training up existing resources. But is this really practical for enterprises and—if so—how can it best be implemented?

Practical Training for Machine Learning Success

If your company has already invested heavily in the app economy and digital transformation, then it will have a good number of employees with technical skills, who could potentially be given targeted and practical training in machine learning. Knowing the problems associated with hiring and outsourcing, this would seem like a great option for enterprises pursuing digital business strategies.

This is complicated by a steep learning curve. Any given approach to any given problem will likely be both praised and condemned by multiple competing academic sources. Parsing all this information is several years work. But remember, the goal is not to become an expert as much as it is to become a skilled practitioner. In other words, it’s possible to bypass much of this academic controversy.

There are certain core algorithms that can be used reliably and with a low barrier to entry. A good way to start, therefore, might be to identify the core algorithms that are most relevant to your use case then identify internal resources who might be well-placed to learn and implement those algorithms. Doubtless, challenges will arise along the way but the popularity of training as an alternative to acquisition, hiring or outsourcing, is clearly well-founded.

Christopher Bonnell
By Christopher Bonnell | March 14, 2018