What is AI? Artificial Intelligence is a term that has been around since the 1950s. For much of that time, it has been a far off fantasy, something only attainable in the future. Now we are right up on the edge of AI’s large scale implementation, but there’s an important question to ask: Is AI an opportunity for breakthrough or a reason for caution?
Look no further than the missteps that AI is already experiencing.
Yes, AI can produce amazing results faster and more accurately, but it is only as accurate as it is trained to be. In July 2018, documents surfaced that IBM’s Watson for Oncology, an AI designed to cure cancer, was making egregious and sometimes dangerous errors in patient recommendations.
The AI had been trained on hypotheticals (a not uncommon practice for AI as actual data collection can be time and capital intensive or difficult for other reasons, such as patient confidentiality) and not real patient data. They had developed an AI algorithm that would cure cancer, but they still treated it like the other AI they knew. They trained this AI like they would train other AI, without realizing this algorithm was a possibility for something new and different. A more abstract way to say this is that when we turn a possibility or act of creation into a “thing,” we stop dealing with it as it is and start dealing with it as we know it “should” be. We lose power in dealing with it – sometimes with disastrous results.
There is no doubt that AI is a technology of the future that will have far-reaching implications and can be utilized to the great benefit of society. However, as AI evolves we need to ensure that we don’t turn AI into a fixed solution. When anything becomes a fixed entity, future possibilities for it are narrowed. The solution to this narrowing is simple, we just need to remind ourselves to keep asking the question: What is AI?