2018 has brought the advancement and broad usage of AI with Voice UI assistants (Alexa, Cortana, Siri), better weather prediction, better agricultural yields and compute that can outperform and these are just some of them.
Dario expects that humans will get complementary expertise as we continue making decisions in the future. Today, institutions such as companies create networks of experts that collaboratively work toward a given goal to solve problems. IBM hopes to have a form of AI as a part of this expert network to support and enrich decision making.
IBM is helping making AI easier to use in technologies. An example is of a transfer learning and bringing reasoning within learning advancements can help reduce the necessary amount of data required. This will directly drive down some of the storage, and data processing, training and inferencing costs.
Automation of AI
Another area where we are likely to see advancements in 2019 and beyond is the automation of AI. It is becoming easier and easier to create machines that automatically construct both the neural network and increase the amount of data science associated with training the system.
The computation for AI is improving dramatically, so you can train more efficiently at lower cost and faster. Increasing the democratization and ability to create AI and ability to deploy AI.
The pillar of 2019 AI advancement will evolve about trusted AI, not just accurate AI as was the case in the past. Learning is not enough for future AI, it will all evolve around reasoning as it allows us to create models of the world and think about possibilities.
The goal is to make AI system learn the way we do.