What are the barriers separating machines from us?
Repeatedly we are seeing artificial intelligence (AI) mentioned on almost every tech website – some voicing their concerns and others looking toward future possibilities. Today, we ourselves are already using different kinds of AI – such as virtual personal assistants, on our phones and on our computers.
Although it may be assumed by many as a recent technology, artificial intelligence is far from being new, and has been studied since the 1950s. So what has made AI important again? Or are we just encountering another cycle of hype like we did in the 1970s?
Artificial intelligence is about creating machines that have the ability to think like the human mind, and being able to do the right thing at the right time. Tech giants are applying AI to all kinds of data, and it is currently being used for smart cars, video games, purchase prediction and fraud prevention.
More recently we are seeing AI being tested in order to create self-driving vehicles, and there have even been claims that IKEA are diving in to the world of AI. From Facebook’s AI claiming to spot suicidal users to AI being able to predict when the heart will fail, the opportunities seem endless.
It’s easy to imagine a machine like Apple’s SIRI or Amazon’s ALEXA in the future engaging with others, answering questions or satisfying commands, however just because they can recognise voice and images, there is still a lot of work to be done. If you watch this video without being critical, it does seem believable.
What are the limitations of artificial intelligence?
Natural intelligence is, by definition, embodied. Artificial intelligence are not yet capable of working the same way that brains do. The reality is that artificial intelligence lacks the ability to understand, let alone answer questions that we might ask to others.
In reality, conversation involves people making assessments of each other and knowing what to say based on their own experience. Some questions require an understanding about different contexts and how people operate in daily life.
Imagine asking an AI programmed machine “I’m going to bake a pie tonight, what do you think?”. Most tools will offer a wealth of recipes, instructions etc., but it will not however tell you what it thinks about that.
AI limitations include the fact that there are many questions that AI are incapable of answering, which can be answered by a human ever so easily. Researches have continuously studied AI’s capabilities over time, and have not yet found a solution for this.
Google’s self-driving car (Source: Smoothgroover22 via Flickr)
AI based search methods are never guaranteed to reach the optimal solution. When using AI in order to resolve an issue, it is difficult to gain true insight into the nature of the problem. AI’s can be referred to as little black boxes that simply seek to map a relationship between output and input variables grounded on a training data set.
This raises multiple questions concerning the ability of the tool to establish situations that were not represented in the data set.
Another limitation, which is not quite a technical limitation, but an issue that needs to be addressed – if artificial intelligence was utilised to build vehicles, who should be responsible if the vehicle were to crash?
Understanding these limitations are crucial to understanding the future potential of AI, and what it means to be “intelligent”. While researchers face multiple problems facing the future of AI, it is not to say that future predictions are unachievable.
The BBC claims that if AI is to be successful, 2050 may be the year that we see it.
Featured image: A health blog via Flickr.