#科技头条#【新一代机器人:通过观察便可模仿人类】

#科技头条#【新一代机器人:通过观察便可模仿人类】日前,谢菲尔德大学研究小组研发了新系统,该系统基于阿兰·图灵1949年的测试,使机器人通过观看就能了解自然系统和人工系统如何运行。研究人员控制了两组机器人,并追踪其行动,若发现不同,学习组便会调整算法以增加和原始组的相似性。这种方法可以用来创建机器人,不用被告知,便可像人类一样行为。

Watch and learn: A new generation of creepy robots could imitate humans simply by observing how we behave

37C07DFA00000578-3765629-image-m-43_1472642409736

A new breed of robots could someday learn to act human without being told how to behave.

These machines are able to learn both how natural and artificial systems work by simply watching them.

This could mean advances in the world of robotics with machines able to predict, among other things, human behaviour, as well as imitate it.

The discovery is inspired by the work of Nazi-codebreaker, Alan Turing, who in 1949, proposed a test in which a machine could pass if its behaviour was indistinguishable from a human.

Dr Roderich Gross from the Department of Automatic Control and Systems Engineering at the University of Sheffield, said: 'Our study uses the Turing test to reveal how a given system - not necessarily a human - works.'

The researchers put a swarm of robots under surveillance, and wanted to find out the rules which governed their movements.

They also put a second swarm of learning robots under surveillance and tracked their movements.

Unlike the original Turing test in which the interrogator trying to distinguish between the two groups was a human, a computer programme, that can learn itself, was given the task.

Dr Gross told MailOnline: 'In our view, machines are better placed for the interrogator's job.

'They may spot subtle differences that we humans may not notice.

37B9336400000578-0-image-a-22_1472575776271

'In a sense one could say "Machines may fool humans, but will they fool machines?"

'This is why Turing Learning is potentially a very powerful paradigm for creating human-level intelligence.'

The machine was tasked with distinguishing between robots from either swarm - the original or the learning.

The computer was rewarded with bitcoins if it correctly identified the correct group of robots.

37B9336000000578-3765629-In_the_experiment_both_the_original_swarm_of_robots_and_a_learni-a-28_1472577809619

But the group of learning robots were also rewarded if they could fool the interrogator into believing their motion was genuine.

If the computer was able to tell the difference between the two, the learning robots then adjusted their algorithm to improve their similarity to the original swarm.

The advantage of this approach – which they have called 'Turing Learning' – is that humans no longer need to tell machines what to look for.

Dr Gross said: 'Imagine you want a robot to paint like Picasso.

'Conventional machine learning algorithms would rate the robot's paintings for how closely they resembled a Picasso.

'But someone would have to tell the algorithms what is considered similar to a Picasso to begin with.

'Turing Learning does not require such prior knowledge. It would simply reward the robot if it painted something that was considered genuine by the interrogators.'

As well as painting like Picasso, the researchers believe Turing Learning could also lead to advances in science and technology.

Dr Gross added: 'Scientists could use it to discover the rules governing natural or artificial systems, especially where behaviour cannot be easily characterised using similarity metrics.'

'Computer games, for example, could gain in realism as virtual players could observe and assume characteristic traits of their human counterparts.

'They would not simply copy the observed behaviour, but rather reveal what makes human players distinctive from the rest.'

Sotheby's Impressionist and Modern Art Sale Feb 2013… Pablo Picasso £25-30M

Sotheby's Impressionist and Modern Art Sale Feb 2013… Pablo Picasso £25-30M

However, Dr Gross added: 'While this should work in theory, it has yet to be seen whether Turing Learning can actually produce Cubism-like paintings in practice.'

So far, Turing Learning has only been tested on robot swarms.

However, the next step is to test whether the machine can learn the workings of some groups of animals, such as schools of fish or colonies of bees.

The researchers hope this could lead to better understanding of the factors that influence the behaviour of these animals.

http://www.dailymail.co.uk/sciencetech/article-3765629/Watch-learn-new-generation-creepy-robots-imitate-humans-simply-observing-behave.html


Comments are closed.



无觅相关文章插件