The best thing about predicting trends is watching the year unfold and seeing which ones really take off. While we have no doubt that 2019 has some jaw-dropping tech surprises in store for us, we couldn’t resist getting the team together to compile some of our best guesses.  

There was a lot of buzz around Natural Language Processing and virtual assistants last year. The ability for AI to process complex images within the health sector – x-rays, for example – really took off. While these advancements are likely to continue into 2019, we believe this year will be all about the engineer: helping the behind-the-scenes work that goes into creating Artificial Intelligence. This year’s biggest trends will revolve around giving data scientists and machine learning engineers better tools for bringing their work to life.

 

2019 AI Industry trends we see in our tealeaves:

Automation

2019 will see us developing tools that automate modelling processes for AI. This essentially boils down to automating automation – and making the whole process of storing, processing and delivering data more efficient for the people who deploy machine learning models, which really makes our brains smile. Watch this space for Konstellation, our T7 evolution, a forthcoming system dedicated to just that.

 

Hardware improvements

With the massive growth of AI comes a need for greater, faster automation – but also processing power. Artificial Intelligence requires a tremendous amount of bandwidth between CPU, GPU and memory, which often creates a bottleneck. Manufacturers are aware of this, and know that AI companies are concerned with delivering a faster time to market. We believe they’ll spend at least a portion of this year creating new processor models that are specially designed for AI projects. We can expect a new Moore’s Law happening on this side in a similar way to the CPU power in the last decades.

 

Adopting DataOps

Applying DevOps practices to Artificial Intelligence? Yes, please. Many AI companies are applying their DevOps mindset to AI projects, and Intelygenz are right there with them. We fully embrace DevOps culture, so it’s been a natural shift for us to apply it within the Terminus7 AI team. Some are calling it DataOps, others AIOps – either way, we love any automated, process-oriented methodology that boosts the quality of our work.

 

Working with non-annotated data

No data? No problem. Where much of Artificial Intelligence has relied on data quality and quantity until now, 2019 will be the year we break free from many of those constraints. Learning by reinforcement – or rather, the ability to train models in spite of not having any annotated data – or working with simulated data to bootstrap initial versions of models will both blossom in 2019, and we couldn’t be more excited about the possibilities that advancement will open up.