In 2011, Guilio Tononi and Christof Koch wrote this article for the Scientific American. Their goal was to be able to provide the lay person a way to test whether a machine is conscious or not. To do this, the authors say that one should only ask the machine a question that any 6-year old child will be able to answer: what is wrong with this picture?
Tononi and Koch both believe that this is a question that a machine will not be able to answer because it would require contextual knowledge – knowledge that can only be gained through experience.
Why It Matters
Six year old children will be able to answer the question, “What is wrong with this picture” because they have some experience of the world and can immediately make sense of what they are seeing. Meanwhile, machines only accept information, they may be able to infer answers to the question, but they will not be able to explain their own response.
This means that there is so much intelligence in the human mind than can be copied by a robot. At this point in time, machines cannot yet surpass this kind of intelligence, but this does not mean that we should be complacent.
New AI is being developed, many of which are bound to be released to the public unregulated and untested for safety. Though we want to trust AI for their super human abilities, the reality is that their current intelligence match only that of rats. Why then do we rely on automatic decision systems based on this kind (and level) of intelligence? They can be expected to be biased, or to make decisions that does not consider the overall impact of its solutions.
But computers are improving periodically, at an ever-increasing pace. Already, machines are already surpassing human capacities in various fields. Sooner or later, AI will catch up. AI may have the capacity for consciousness in the next few years, or at the very least, act as if it is conscious. It may even be able to pass the test for consciousness suggested by this article. But until the alignment challenge remains unaddressed, we hope that no conscious, or general intelligence (AGI) machines will be built. Otherwise, we might end up as the lab rats of our own creations.Testing-for-Consciousness-in-Machines