Playing chess, driving and even composing music – for many decades, skeptics have argued that these skills will never be mastered by robots with artificial intelligence (AI). Today, one hears fundamentally different projections. Futurologist Yury Vizilter, department head at State Research Institute of Aviation Systems, says functional artificial intelligence will be almost complete by 2020: robots capable of performing tasks that remain the privilege of humans are a footstep away from reality. On the other hand, a few skills are still beyond smart machines: the Digital Trends technology portal has counted at least six. Some sound kind of lightweight, but they clearly show that machine intelligence is still far from fully replacing a person. Invest Foresight offers a brief description of each of them.
No laughing matter
Coming up with a really funny joke is not the easiest task for a person. It is not surprising that all attempts to generate humor with the help of artificial intelligence have ended in failure. In early 2018, US researcher Janelle Shane tried to teach a neural network humor using a database of 43,000 ‘what if’ jokes. It would seem that the system had every chance to learn how to be funny. However, the jokes that the machine tirelessly produced could hardly elicit applause. Humor remained inaccessible to neural networks, even though artificial intelligence has repeatedly demonstrated the ability to understand complex linguistic structures.
Romancing the network
Smart systems with artificial intelligence are quite capable of generating coherent texts. Yet, these ‘authors’ are unlikely to produce a novel good enough for reading – a popular paperback love story or a piece of high artistic merit. Obviously, books require a bit more than the ability to write a coherent text. The author at least has to come up with an idea that would interest the reader and choose the best way to convey it to the audience. True, there are several projects using AI for writing prose – some successful, some not so good. However, no one even expects robots to produce texts comparable to novels by Jane Austen or JK Rowling.
Creativity not their cup of tea
Artificial intelligence systems have already learnt building optimal strategies. And of course, creativity is not a trait unique to humans. Yet, when it comes to creative strategies developed by a lawyer who picks the best arguments to defend the client, or by a company chief choosing targets for the company’s development, AI systems will be of little use. In this case, this is not only about analyzing the data but about being ready to tackle unstructured tasks and decide which information is important and which can be disregarded.
Have a heart, robot!
This is not about humanization in a literal sense, but rather about such concepts as compassion, typical of humans. If robots could diagnose cancer, who would one choose to hear the bad news from – a human doctor or a smart machine? Indeed, Big Data analysis is successfully used in sports via neural networks – but can AI perform the work of a coach? These are important human roles, and they will continue to remain such in the foreseeable future.
Can I have a cup of coffee?
You can easily find a smart coffee machine in a supermarket, but will a robot pass the so called ‘coffee test’? The test was suggested by Apple co-founder Steve Wozniak as an assessment of various aspects of AI. In this test, a robot would be challenged to perform a very simple task – to enter your home, find the coffee machine, and brew a cup of coffee by pushing the right buttons.
This test makes AI capabilities visually measurable, unlike other attempts at quantitative assessments of a system’s capabilities (such as the Turing test, which focuses on philosophical abstractions). Obviously, AI cannot yet cope with the ‘coffee’ task, which requires a generalized multipurpose intellect. This condition is not easily achievable so far.
Playing in a team
Artificial intelligence is ready to secure victories by beating the world’s top Go and chess players, but can similar achievement be expected from robots? To successfully compete with people, they have to truly become multipurpose. The task has been partially solved by Boston Dynamics developers, whose robot excellently performs sport exercises and has even leaned parkour. But will we ever see a team of robots defeating a human football team? Given the speed and numerous skills this sport requires, the task seems achievable only in the extremely far future.
By Olga Blinova