This week - DeepMind's new AI get closer to general-purpose AI; Microsoft's patent to create chatbots from anyone, including dead; a four-legged robot with wheels; and more!
More Than A Human
Microsoft has filed a patent for chatbots that can mimic anyone if supplied with enough personal data. The patent describes how using a collection of data about someone (posts, photos, videos, comments, voice recordings) can be used to improve its customer service chatbots, as well as future AI assistants. As this article points out, the same technology can be used to pretend to be someone, even someone who is dead, raising questions about privacy after death.
DeepMind presents MuZero, "a significant step forward in the pursuit of general-purpose algorithms" as the company describes it. MuZero was able to reach the state the art result on the Atari benchmark, while simultaneously matching the performance of AlphaZero in the classic planning challenges of Go, chess and shogi. DeepMind hopes this AI will be useful in solving new challenges in robotics, industrial systems and other messy real-world environments where the “rules of the game” are not known.
"Replication problems plague the field of AI, and the goal of general intelligence remains as elusive as ever", read the headline of this article. The replication part refers to companies claiming their AI outperforming humans in any activity but when tested, it fails to replicate the results (the article uses Google Health as an example). It gives other examples of areas where AI does not live up to its hype and quotes Gary Marcus' warning that “if and when the public, governments, and investment community recognize that they have been sold an unrealistic picture of AI’s strengths and weaknesses that doesn't match reality, a new AI winter may commence.”
Interesting observation - companies realised their corporate filings are most likely to be read by machines so they are optimising them to be not only machine-friendly but also to manipulate the machine to draw favourable conclusions about the content, like avoiding words that are listed as negative or adjusting the tones of voice used in the standard quarterly conference calls with analysts, because they suspect those on the other end of the call are using voice analysis software to identify vocal patterns and emotions in their commentary.
In the discussion of wheels vs legs for robots, Swiss company ANYmal said "why not both" and showed how this hybrid solution of attaching wheels to their four-legged robot can improve robot's mobility.
US Army is interested in biohybrid robotics - robots that use muscle tissue to produce never-seen-before agility and versatility. The first applications for biohybrid robotics the team expects to focus on are legged, similar to Boston Dynamic's BigDog but researchers are also considering flapping-wing drones.
A team of researchers is creating mobile robots for military applications that can determine, with or without human intervention, whether wheels or legs are more suitable to travel across terrains. The team is developing an adaptable Wheel-and-Leg Transformable Robot (α-WaLTR) that can traverse over varying surfaces, including staircases, more efficiently. The α-WaLTR will move with wheels or legs depending on their immediate need and will be able to decide for itself which to use.
When imagining the city of the future, architects, designers and artists were putting technology in the centre of their designs. Recently, however, there is a shift from a technology-focused approach to reimagining cities to asking how to design a city that addresses the inequality and environmental sustainability problems.