The State Of AI in 2019: Breakthroughs In Machine Learning, Natural Language Processing, Games, Deep Learning & Domain Knowledge

The State Of AI in 2019: Breakthroughs In Machine Learning, Natural Language Processing, Games, Deep Learning & Domain Knowledge

                     Today if we talk about the most rapidly growing domains then the first name that comes into our mind is AI. AI records and stocks are kept not only with the help of constant attention but also with the ability to dissect and evaluate along different dimensions. The founder of Air Street Capital and RAAIS  Nathan Benaich and also AI angel investor has already chosen this way. The State Of AI Report which was published on June 28. Benaich and Hogarth embark on a 136 slide long journey on all of the things. The parameters include AI: technology breakthroughs and their capabilities, supply, demand and the concentration of talent working in the field. Benaich and Hogarth are not just the venture capitalists but they both have extensive AI backgrounds. They have already worked on many AI initiatives from researches to the startups. Moreover, they draw their expertise of prominent figures like Google AI Researcher and lead of Keras Deep Learning Framework Francois Chollet, VC and AI thought leader Kai-Fu Lee and Facebook AI Researcher Sebastian Riedel. This work includes the efforts of all rich expertise.

Unpacking AI

If you have already worked on AI then this is definitely not the first AI report you have noticed. There are many people who are familiar with FirstMark’s Data and AI landscape which is compiled by Matt Truck and Lisa Xu. They must also be familiar with The State of Divergence by MMC Ventures. The updates of all these reports were reached simultaneously. Though there is an overlap but also differentiation in the terms of content as well as approach and format.  The First Mark’s Report has listed players ranging from data infrastructure to AI quite extensively. While on the other hand, the Big Data to AI evolution is a natural one. Thirdly the MMC ventures have a different point of view and have become more abstract.

State Of AI Report 2019

The State Of AI in 2019: Breakthroughs In Machine Learning, Natural Language Processing, Games, Deep Learning & Domain Knowledge

Benaich was asked a question that what is the most valuable knowledge among what they share? He answered that they believe that AI will prove as a force multiplier on technological progress in increasing digital and data-driven world.  This is because everything that surrounds us from cultural to consumer products is a product of intelligence.  The report matches Benaich’s goals and the first 40 pages focus on the progress of AI’s research which includes technology breakthroughs and capabilities. Key areas are reinforcement learning, applications in games and future directions, natural language processing breakthroughs, deep learning in medicine and Auto ML.

Reinforcement Learning, Games and Learning in the Real World 

It is an area of machine learning that is worked on by the researchers over the past decades. Benaich and Hogarth define it saying it as the software agents that learn goal-oriented behavior by trial and error in an environment that provides rewards or penalties in response to the agent’s action towards achieving the goal. The progress in RL is made with training AI to play games which are equating or surpassing human performance. StarCraft II. Quake III Arena and Montezuma’s revenge are just some of those games. The sensational aspect is that AI beats humans however these are the methods which have resulted in play driven learning, simulation, and real-world combination and also curiosity-driven exploration.

Simulation is used by open AI to train a robot which shuffles physical objects with impressive dexterity. Moreover, in RL, the agents learn tasks using trial and error method.

Natural Language Processing and Common Sense Reasoning

It has been a big year in natural language processing(NLP): Google’s AI’s BERT and Transformer: Allen Institute’s ELMo and others have also demonstrated that pre-trained language models can substantially improve their performances on tasks. Pre-training models are to learn high and low-level features that have been transformative in computer vision.

Combining Deep Learning and Domain Knowledge

Benaich was questioned for his taking approaches to combine deep learning and domain knowledge for NLP. This is something that experts think is a promising direction. He also concluded saying that combining deep knowledge with domain knowledge is a fruitful avenue of exploration.

Spread the love

You might like

Erin Phan

About the Author: Erin Phan