What’s the state of AI in 2020?

Although expected with a lot of hype by many, since it marks the beginning of a new decade, 2020 has been quite controversial year so far. Kicking off with rampant wildfires in Australia and continuing with a global pandemic that forced everyone to adapt to a new way of living, most can say that this year hasn’t been one of the best so far. However, “the show must go on”, no matter the circumstances, and today we will look on the bright side of things, the most intresting innovations from the field of AI that took place this year so far: 

Natural Language Processing 

The hottest area in AI at the moment seems to be NLP (Natural Language Processing), which is dominated by large companies and huge models with massive training costs. If the most complex model developed last year (T5-11B) had 11 billion parameters, the most advanced model from 2020 (the GPT-3) has raised the bar to 175 billion parameters. The costs required for training such a model are assumed to be around 10 million dollars.

Deep Learning

Deep Learning continues to get more efficient as it consumes more data. Since 2012 the amount of computing power needed to train a neural network to the same performance on ImageNet classification has been decreasing by a factor of 2 every 16 months. 

Translation of Programming Languages

An unsupervised machine translation model trained with 1000 parallel functions can translate 90% of these functions from C++ to Java and 57% of Python functions into C++ and successfully pass unit tests. 

Code Repairer

Stanford University developed a tool (DrRepair) that given a broken program and diagnostic feedback (compiler error message), localizes an erroneous line of code and generates a repaired one. 

More Students Enrol to AI Courses

The Eindhoven University of Technology (TUE) will spend 100 million euros over the next 5 years to create a new institute focusing on the use of smart algorithms in machines like robots and autonomous cars. Abu Dhabi opens the world’s first AI university (Mohamed bin Zayed University of AI or MBUZAI), which received 2223 applicants of 97 nationalities, coming from 31 countries. Leading universities continue to expand the AI course enrollment: Stanford teaches twice as many students per year as in 2012-2014. 

AI-designed Drugs

Biology is experiencing its “AI moment”, as the number of publications involving AI methods in biology are growing with more than 50% each year since 2017. Papers published since 2019 account for 25%of all output since 2000. The first phase 1 clinical trial of an AI-designed drug begins in Japan to treat OCD patients. Deep Learning models that interpret protein biology are used to find new therapeutics. 

Robotic Process Automation

Robotic Process Automation (RPA) and Computer Vision (CV) are the most deployed technique in industry. Over half of all US states have enacted legislation related to autonomous vehicles, AI dialogue assistants are live and handling calls from UK customers right at this moment and more than 50% of the world allows the use of facial recognition at the moment.

And the years is not even over yet! What are some other remarkable innovations from the field which are not listed here? Let us know in the comment section.