How Computers Think

How powerful can AI be?


In a previous article we came to the conclusion that it is not unrealistic to expect robots to enter the consumer market by the second half of this century.

Powered by the development of Artificial Intelligence, the robot that we saw Theodore fall in love with in the film Her, although not on the horizon, may not be as far away as we all thought.

Many assume that the increasing sophistication of AI will only have an effect on scientists and companies who are looking at smarter ways to do manufacturing. The truth however is that over the next fifty-years, we could be surrounded by machines that have the capacity to rival our level of intelligence.

Whilst that remains a distinct possibility for the future, we’re still some way off mixing with robots that can communicate with us freely. There have however been some interesting developments in the field of late, some of which could be viewed as a step forward.

For example, gaming researches have developed an ‘intelligent’ Mario which uses cognitive modelling to allow the character to move with rational and as if here were a real person. Although some commentators have suggested that this allows Mario to think freely and subjectively, he’s actually a bot that’s programmed to the nth degree. This makes the Mario experiment interesting, but not a massive step forward in terms of coming up with a free-thinking machine.

One of the main issues with the AI-powered examples is that at the moment they are only good at one thing, with that normally either vision or sound. These systems are often far from singular and fail to combine information in a way which is normal for a human.

It comes down to the fact that although AI has allowed these systems to understand and recognise people, they haven’t found a way of getting them to think on a philosophical level, which in essence, is what makes us human.

Neural networks are interesting because they teach themselves new things as they see more data, just like we do when we’re revising for an exam or just reading the news. But at the same time, these systems are not versatile, they are trained to be good at recognising speech or vision meaning that they often fail to pick up on certain things.

As it stands, AI has yet to bring us to a point where we’re being introduced to machines that have the ability to communicate on a human level. We will get there, but today’s systems remain rudimentary.


Read next:

Panel Session: The Cultural Shift