The development of artificial intelligence (AI) is accelerating at a rate faster than most other technology at the moment. Most of this is due to the huge investments that are being made in the field. As with everything, AI also has trends and these are 5 of the most interesting ones to watch out for this forthcoming year:
Decentralised AI: Rather than training machine learning (ML) models in a centralised manner, AI companies are now starting to decentralise things. The way this has been approached so far can be seen in Google’s Android keyboard.
It allows users to effectively train ML models on Android devices with their user data. Making AI training available on mobile devices could help solve many of the problems companies face when using centralised learning.
On-device AI/Core ML: CoreML is a foundational ML framework that has is helping to make the move towards on-device AI. It can easily be integrated into iOS applications and is looking to make AI and ML apps available to the public in 2018.
AI content creation: Using AI to create readable and engaging content is a trend that will definitely be continuing this year. With advancements being made in natural language processing (NLP) and natural language generation (NLG) companies are now able to create readable content in a matter of seconds. AI content generation is already being used social media, news, financial reporting, and marketing and is only going to get more popular.
AI offline: More often than not, AI-based solutions are there to help solve online issues. However, companies such as Amazon are starting to more towards offline data as well. What this essentially means is collecting data via sensors and actuators in malls and stores. These types of devices can already be seen in Amazon Go grocery stores and are a great way to track customer’s movements and shopping habits.
The rise of capsule networks (CapsNet): Proposed by one of Google’s lead scientist’s, CapsNet is a new form of deep neural network that looks to overcome the failings of convolutional neural networks (CNNs).
While CNNs are good at recognizing images similar to those fed to it during training, they have poor performance if presented with anything slightly out. Capsule networks, on the other hand, can recognize objects and images no matter how they’re positioned and are set to be the next big thing in computer vision and image recognition.