AI & Deep Learning: 2019 Predictions

Wouldn’t it be great if AI could tell us what 2019 had in store? After all, machines are smarter than humans, right? Weren’t robots supposed to take over the world already, and AGI have been reached?

Not so fast.

Whilst we’ve still made some incredible progressions over the past 12 months, we’re very far off a world where machine intelligence is superior to that of humans. After all, I asked Siri this morning if I should take a jacket - the response? ‘I can’t find a number for Tekken Jacket’. We’re wise not to panic about a robot-dominated society just yet.

This year, however, we’ve seen breakthroughs in medical imaging, with NLP, with AI cloud-based services amongst others. So what does 2019 have in store? We spoke with some global AI experts to hear what’s top of their prediction list to see what the new year will bring for AI. We heard about real-world practical applications, more cloud-based services, and people becoming more reliant on AI for positive social impact.

Here’s what the experts have to say:

Jeff Clune, Uber AI Labs

We very recently showed that our new Go-Explore algorithm can solve previously unsolvable “hard-exploration problems”, specifically the benchmark challenges of Montezuma’s Revenge and Pitfall (two of the hardest Atari games for AI). I predict many will apply Go-Explore to a variety of domains that require exploration and find that it makes dramatic improvements on them. The first results will come across many video games, but the most interesting progress will be in robotics. Specifically, I predict that Go-Explore will dramatically improve the ability of robots to learn to do very complex tasks automatically.

Mark Weber, IBM

We will see lots more work on Graph Convolutional Networks - scaling, dealing with time series and dynamism - and experimenting with use cases in finance, pharma, genetics, and other problems where graph analytics are helpful. There will also be more work on machine learning for causal inference to augment humans with deeper insight into complex system dynamics, as well as a spike of work on ODE solvers as a new alternative to neural networks; see basic explanation here.

Aditya Kaul, Tractica

Reinforcement Learning will see increasing adoption in the enterprise breaking out of silos which have limited it to simulation-based environments into AI model production environments like Facebook’s Horizon and expanding into other frameworks. Expect to see breakthroughs in meta-learning which allows AI models to oversee other AI models in learning better, especially focused on finding better neural network architectures using Neuroevolution, which should pave the way for the first step of introducing meta-learning into AI frameworks. 2019 will see rapid progress in the first AI-enabled drug discovery candidates, which will be further accelerated by AI-based clinical trials, with the first AI-enabled drugs becoming available by 2020-21. Autonomous retail will gain traction in 2019 as traditional grocery chains push for an accelerated strategy of autonomous pop-up stores across major cities in the US and Europe in response to Amazon’s own strategy. 2019 will see the first large-scale AI-enabled cyber attack either on an enterprise or a nation-state.

Fiona McEvoy, YouTheData.com

In 2019 we can expect conversations around AI and tech ethics to intensify, especially when it comes to consumer data and privacy. We may also see the field widen its gaze to less mature technologies, like augmented reality. Importantly, more companies will be trying to anticipate how their tech could negatively affect us in the future and making efforts to bake ethical principles into product development as standard.

Hao Yi Ong, Lyft

I think folks will continue to see more exciting industry real-world use-cases in 2019. Key 2018 product rollouts such as AWS, Azure, and GCP's powerful GPU cloud instances and investments in easing the deployment of GPUs in data centres such as Nvidia support for Kubernetes are setting the stage for an exciting year ahead. 2019 will see the rise and acceleration of deep learning applications in industry by smaller players fueled by improved access to powerful compute resources previously available only to bigger players.

Keith Adams, Slack

Controversially I think C++ coders are going to be replaced long before truck drivers. I think autonomous vehicles are at the far end of AI. If a car needs to know how to drive in India, it needs to understand so many things about the road, like How an elephant moves in traffic!

What do you think will be the hottest trends of this year? Let us know in the comments. If you're keen to stay ahead of the curve this year, join us at the Deep Learning Summit in San Francisco this January 24 - 25 to learn from some of the leading minds in AI & DL such as Ian Goodfellow, Ilya Sutskever, Chelsea Finn & many more. Register with the code NEWYEAR before Jan 11tth to save 25% on all passes.