A Year In Artificial Intelligence: 2018’s Key AI Development Milestones

A Year In Artificial Intelligence: 2018’s Key AI Development Milestones

2018 is drawing to a close. It’s oft repeated how much developments in AI and robotics are now accelerating after a couple of decades of buzz and little notable practical application for the latest technology in the sector. The emergence of cloud computing has meant the compilation of, storage and easy access to big data has taken a huge leap forward.

Big data has become both practical and cost effective and is the fuel feeding the learning of AI algorithms. And faster, cheaper more efficient processors like CPUs and the latest generations of microchips that are improving even on that technology mean algorithms can assimilate and learn from those huge reservoirs of data at ever greater speeds.

AI in all of its forms, such as the neural network structures of Deep Learning, has reached a tipping point and is not beginning to realise its potential and genuinely make an impact on what the latest technology in the world can achieve. However, we are also still a long way away from what AI-powered technology will be able to achieve in the future across finance, medicine, construction, manufacturing, home help, entertainment and pretty much any other sector you might care to mention. But 2018 has been a year of significant AI progress. Looking back at the year gone by, what have been some of the most notable AI landmarks?

AI-Powered Mobile Robotics Performing Human Jobs

Programmed robots being able to take on jobs previously performed by human workers is of course not a brand new development. Production lines have been gradually incorporating more and more robotics for at least a couple of decades now. However, AI-powered mobile robotics able to navigate their way around complex environments and perform jobs that require an assessment of different factors and context are new.

This year a quadruped robot, inspired by canine biology, spent one week performing inspections on one of the biggest offshore power-distribution platforms in the world. The ANYmal, developed by Swiss start-up ANYbotics, was given a trial checking the platform’s gauges, levers, oil and water levels among other visual and thermal checks across 16 checkpoints. The success of the trial suggests that within a decade much of the off-shore industry’s requirement for human involvement in day-to-day checks and maintenance could be replaced by robots.

Virtual Urban Space Generation/Simulation

Nvidia, the U.S. technology giant best known as the developer of the CPI microchips whose processing power is playing such a significant role in enabling AI advancements, demonstrated how an algorithm its experts have developed was able to generate a detailed 3D generation of a hugely detailed sim-city. The Nvidia algo uses data gathered by smart dash cams to generate its city simulations. Practical applications for the technology are potentially saving computer game publishers a huge amount of resources in the hours needed to programme virtual environments as well as training driverless car technology in virtual environments.

In the future, we can expect new towns and cities to be planned to optimal efficiency using future generations of similar technology.

AI Produces Artwork That Sells for Six Figures

Whether it was the aesthetic and meaning imbued itself, the novelty of the fact that it was produced by AI or an attempt to curry favour in the event of a future robot uprising, 2018 saw a painting produced by AI sold at auction for almost half a million dollars. Portrait of Edmund de Belamy, a portrayal of a well-fed-looking man in black frock jacket and white frilly collar went to a winning bid of $432,000 at an auction hosted by Christie’s.

The significance of the development is that AI is most associated with tasks based on patterns being found in numbers and other data rather than creativity. The arts are considered to be the strongest bastion of humanity and AI encroaching into even this realm gives pause for thought.

AI Holds a Realistic Conversation

2018 saw Google unveil its ‘Duplex’ AI. The Google Duplex, at least on the basis of the showcase of the technology given at the company’s summer Google I/O (internet of things) 2018 event, represents a genuine breakthrough in AI’s ability to maintain a natural-sounding conversation. AI achieved strong levels of voice recognition and the ability to respond to commands and offer responses some time ago but there has never been anything really approaching the illusion of interacting with another human.

The Duplex demonstration included the algorithm taking restaurant reservations being made over the phone, with the caller completely unaware they were not speaking to a human. Even common ‘parasites’ or mechanism’s to gain the time to process thoughts like ‘aaahhm’ and ‘hmmm’ are present in the way Duplex speaks. While still officially a ‘demo’ product, the Duplex AI demonstrates that it may not be long before jobs such as waiting and other hospitality roles as well as phone-based customer service to name just a few service industry occupations will disappear, performed by AI.

Deliveries Performed by AI-Powered Robots

Starship Technologies, the perhaps aptly named robot manufacturer that has raised millions in investment capital this year started deploying its wheeled delivery bots for real deliveries in a number of cities across the U.S. The bots only performed ‘last mile’ deliveries from a deployment point located a short distance from the last leg of the deliveries but it surely won’t be long before the bots are able to take on deliveries from starting points much further away.

Customers waiting for deliveries being made by the Starship Technologies bots also have the convenience of being able to say exactly when they want their delivery to be made and to track the bot bringing their parcel in real time through an app.

Deep Fakes Confuse Reality

‘Deep Fake’ technology uses AI to augment real videos by superimposing ‘fake’ faces or voices onto footage. The results are scarily realistic and almost impossible to tell apart from real video content. Among the most well publicised examples of deep fake technology were a video of former U.S. president Barack Obama making aggressively derogatory comments about his successor Donald Trump. In fact, Obama cooperated in the deep fake footage to highlight the risk of how convincing the technology can be and how it could be used maliciously to manipulate minds and opinions in a new, much deeper and more dangerous version of ‘fake news’.

There are many more examples of major AI milestones to have been made this year. Whole articles much longer than this one could be dedicated to covering landmarks in AI applied in specific areas from medicine to agriculture. But the examples provided are a taster of just how fast things are moving when it comes to the latest AI technology in the world. Doubtless we’ll be able to round up 2019 with even more impressive AI developments to have unfolded over the course of the year.

Leave a Comment