Quantum, ML&AI, Robotics

5 min readMar 25, 2018

Note: I wrote this part back in December 2017 for ECFR’s end of year publication on “Opportunities amid disorder: Europe and the world in 2018” (http://www.ecfr.eu/publications/summary/opportunities_amid_disorder_europe_and_the_world_in_2018). While contributing authors were mentioned in a paragraph at the end, the publication does not indicate as to who wrote which parts of the paper. To rectify that, here is my unedited contribution.

Quantum Technology

Hype and reality on the progress of quantum technology will continue to rise and ebb as developments accelerate but remain poorly understood outside the scientific community. In August 2016, the People’s Republic of China captured headlines around the world for launching the world’s first quantum-communications satellite. Named after Chinese philosopher Micius, the satellite proved in July 2017 that particles can remain linked in a quantum state at a distance of more than 1,200 km.[1] Hailed as a major stepping stone toward ‘unbreakable’ encryption, quantum-everything has since become the least understood buzzword.

Progress in the field of quantum technology is indeed slower and faster than many people tend to realize. In November 2017, IBM for instance announced the creation of a 50 qubit prototype quantum computer which managed to keep its qubits in a quantum state for 90 microseconds — which is double the coherence time of IBM’s smaller 20 qubit machine.[2] As MIT Technology reviewer Russ Juskalian explained, “the top supercomputer systems can currently do all the same things that five- to 20-qubit quantum computers can, but at around 50 qubits this becomes physically impossible.”[3] But while we are still at least 5–10 years away from getting quantum computers to work on an industrial scale, both governments and companies are engaged in fierce competition in the quantum field.

Based on the success of the Micius, China announced a plan to create the world’s first unhackable computer network.[4] DARPA meanwhile aims to create the first quantum key distribution network, and IARPA is funding research into logical qubits, coherent superconducting qubits, and error-free quantum computing.[5] Meanwhile in Europe, the Commission’s € 1-billion quantum project is slowly taking shape.[6] With IBM, Google, Microsoft, Intel, and several promising start-ups working energetically in the quantum field, 2018 is guaranteed to witness several technological breakthroughs on our journey into the quantum age.[7]

Machine Learning and Artificial Intelligence

Artificial intelligence and machine learning are intertwined at the hip and naturally progress in lockstep. In many ways it comes down to the difference between Google’s AlphaGo Zero and Microsoft’s Tay. While the Twittersphere turned Tay into a racist, homophobe chatbot within 24 hours of its launch in March 2016,[8] AlphaGo Zero applied reinforced learning for the first time — by playing against itself (neural-network software) and utilizing search algorithms to predict moves — freeing itself from the constrains of human knowledge.[9] In the end, it took AlphaGo Zero just 40 days to surpass all other previous versions of AlphaGo and arguably become the best go player in the world in October 2017.[10]

As advancements on neural-network software are progressing and machine learning techniques are refined or entirely reworked, the dream of creating artificial intelligence is coming ever closer within our grasp. Capsule Networks or CapsNet are the latest trend in artificial intelligence and machine learning to get computer vision right. None other than the godfather of artificial neural networks Geoff Hinton, pushed the idea of capsule networks in two of his papers in November 2017.[11] The idea behind CapsNet is to narrow the gap between AI systems and a human toddler to understand their surroundings. Instead of feeding the system more and more data to comprehend different object from different angles in different positions (positional invariance and equivariance), capsules are designed to track different parts of an object and their relative position in space. Meaning that a network of capsules can build awareness to understand when a new object is actually a known object from a different view.[12]

Given the progress in artificial intelligence and machine learning, it should not come as a surprise that different stakeholders predict vastly different outcomes on the future of AI. While Russian President Vladimir Putin essentially kicked off the AI race between nation states in September 2017, by noting that “whoever becomes the leader in this sphere will become the ruler of the world,”[13] Tesla’s Elon Musk warned that the “competition for AI superiority at national level [will] most likely cause WW3.”[14] Meanwhile, Facebook’s Mark Zuckerberg denounced Elon’s warning as “pretty irresponsible,”[15] and Google’s Head of Search and AI John Giannandrea declared that he is “definitely not worried about the AI apocalypse.”[16] Whatever the military adoption and future of artificial intelligence may be, 2018 will most likely be another critical leap into the unknown.

Robotics

DARPAs Robotics Challenge (2012–2015) was an eye-opening revelation to the robotics community in terms of drama, inspiration, hardships, and fun. Indeed, making robots move semi-autonomously in the field to complete complex handling tasks is an extremely difficult feature to pull off. And while the Challenge could have been even harder with contenders trying to solve even more practical problems (such as how to stand up a robot after it has fallen over), the Robotics Challenge has given the community a sense of the here-and-now, and a reality check on the immense challenges that are still lying ahead.[17]

Given that robotics is a wide field by its very own definition, progress varies across different domains. Take Sophia for instance, the first humanoid robot that was awarded honorary citizenship by Saudi Arabia in late October 2017.[18] Or be awed by Boston Dynamics Atlas, who “leaped from box to box like a gazelle,” did a 180, and finished with a backflip.[19] Atlas’ performance is simply stunning when compared to the numerous fall-over tragedies at DARPA’s Robotics Challenge or Honda’s Asimo falling off a set of stairs 11 years ago. As Paul Miller over at The Verge beautifully put it, “there was a time before Atlas could do backflips, back when robots were for factories, bomb disposal, vacuuming, and the occasional gimmick, and none of the useful ones were humanoids. Now we’re living in an era where humanoid robots are apparently as agile as we are. So what will they be used for? It’s time to get out the popcorn.”[20] And maybe, if French President Macron is able to push through his idea of a European version of DARPA, the next Robotics Challenge could take place on European soil, with European know-how, and European technology.[21]

[1] https://www.nature.com/news/china-s-quantum-satellite-clears-major-hurdle-on-way-to-ultrasecure-communications-1.22142

[2] https://www.ibm.com/blogs/research/2017/11/the-future-is-quantum/

[3] https://www.technologyreview.com/s/603495/10-breakthrough-technologies-2017-practical-quantum-computers/

[4] https://phys.org/news/2017-08-china-world-quantum-network.html

[5] https://www.darpa.mil/about-us/timeline/quantum-key-distribution-network; https://www.iarpa.gov/index.php/research-programs/quantum-programs-at-iarpa

[6] https://www.nature.com/news/europe-s-billion-euro-quantum-project-takes-shape-1.21925

[7] https://www.technologyreview.com/the-download/609484/why-this-new-quantum-computing-startup-has-a-real-shot-at-beating-its/

[8] https://www.theverge.com/2016/3/24/11297050/tay-microsoft-chatbot-racist

[9] https://deepmind.com/blog/alphago-zero-learning-scratch/

[10] https://deepmind.com/blog/alphago-zero-learning-scratch/

[11] https://arxiv.org/abs/1710.09829; https://openreview.net/forum?id=HJWLfGWRb&noteId=HJWLfGWRb

[12] https://www.wired.com/story/googles-ai-wizard-unveils-a-new-twist-on-neural-networks/