5 AI & Data Science Predictions for 2020

2019 was a tremendous year for artificial intelligence (AI) development. Not only were 100 research papers published every single day, but the past year also became a record-breaker for AI funding.

All of this begs the question: What’s in store for AI and data science in 2020? Let’s find out with five predictions!

1) AI Ethics Will Become More Concrete

Combining AI automation with decision making has always caused some concern, and high-profile incidents like Amazon’s scrapped recruiting tool that showed gender bias and the Cambridge Analytica Scandal have only served to stoke these fears in the public consciousness. Fortunately, it has also given some much-needed momentum to data science and AI ethics.

Finally forming into its own defined discipline, data science and AI ethics are now hyper-focused on a few key areas: automated decision making, algorithmic bias, privacy and consent, when to have a human-in-the-loop, and long-term challenges that humanity will likely encounter on its path towards creating artificial general intelligence.

While fleshing out ethics for AI and data science, it’s important to consider how they’ll interact with global privacy regulations. For example, GDPR has been in full effect for almost two years now. This EU regulation places limitations on data processing and profiling, and it requires a certain level of transparency to be maintained. It even raises the possibility for organizations that employ data scientists to be held accountable for any unfavorable consequences.

It’s completely normal for new technology to outpace regulation for a few years. But it seems that regulation is finally catching up to AI and data science. In the short-term, this will cause some discomfort as data scientists and AI engineers learn to work within newly established constraints. But it should pay off in the long-term by weeding out bad actors from the bunch.

2) Too Many AI Tools = Confusion

Today, you can accomplish any AI task in a variety of ways, thanks to the plethora of tools available. Obviously, different people will prefer different approaches. But this will likely continue to cause colossal confusion among newcomers to the field.

For instance, if you work for a large company, you could build a machine learning model using enterprise tools. If you’re a database administrator (DBA), you could also model this in a database environment with Microsoft SQL Server. If you’re a software engineer, you could lean on machine learning APIs to build your AI product. And if you’re more familiar with the cloud, then you could tap into the power of platforms like AWS Sagemaker or Azure ML Studio to get the job done.

Obviously, the possibilities go on and on. But this myriad of options could cause misunderstanding between different groups. Organizations that are able to cultivate meaningful collaboration between their disparate technical teams will end up being the most successful in breaking down these barriers.

3) The Trend Towards Specialization Will Continue

Expect to see specialization to become more prominent for both data science and AI roles. Right now, the roles in these fields are already essentially split into “engineering” and “scientific” roles.

In the former category, we have roles like Data/Machine Learning/AI Engineers. They mostly focus on the infrastructure and platforms that support large production systems. It’s common for those in these positions to have a history or natural affinity with IT. For example, many software engineers have transitioned to a role in this space. Demand for these positions will continue as more AI models move into production.

In the latter category, we have Business Analysts, Data Scientists, and Consultants. These roles tend to focus more on investigative work and decision support. Similar to the engineering roles, we can expect the demand for these positions to continue, as there is no shortage of uncertain situations requiring careful navigation and help with decision making.

So, how are both of these positions continuing to surge in demand? Well, many organizations are quietly moving away from the concept of finding a “unicorn,” or someone who is capable of doing everything. Being multi-talented sounds nice. But unless you’re a scrappy San Francisco development startup or solopreneur, you’re probably better off hiring those who specialize and excel in one distinct trade — especially if you’re striving to build and scale a large team.

4) The Rise of AI Literacy

All of the AI advancements in the world don’t mean anything if their actual consumers can’t comprehend or use them. Many AI industry players have been too busy to realize that this is the ultimate bottleneck to data science value, not the technical aspects and challenges they’re focused on during their day-to-day.

Numerous organizations have gotten a head-start on addressing this issue by offering training for their employees. It’s becoming clear that in-house programs are often the best way to approach this since they allow for content customization. This may appear as a steeper initial investment, but it’s well worth the cost since it allows companies to align training and education with the problems they’re facing and the data they have at hand.

5) End-to-End Management of AI Models

When AI models move into production, it’s best to maintain them with end-to-end management. This has become painfully clear as the size and scope of AI projects have increased immensely over the past few years. In fact, this growth has formed the foundation for end-to-end model management to become its own discipline in AI. This includes but is not limited to model deployment, model monitoring, different support tiers, and identifying when to retrain or rebuild models.

Model monitoring, or Model Ops as it’s called in the field, has grown into its own distinct set of skills that differs from those of the scientist and engineering roles we discussed earlier in this post. As a result, this niche is shaping both of those roles and their evolution.

Stay Tuned!

We hope you’ve enjoyed this fun forecast of what’s to come for AI in 2020! Stay tuned, because we’ve actually been saving a special surprise: We have five more predictions for AI this year!

Which prediction from this list do you think will happen first? Which prediction do you think won’t come true at all? Let us know your thoughts in the comments below!