Artificial Intelligence is becoming somewhat mainstream. I say somewhat, because it’s probably not being used in our daily lives just yet. That being said, you would be surprised where it exists. Google Photos, for examples is able to identify pictures of mountains, using AI. Thus providing context to your images and giving you the ability to search them. Maybe you already knew this, but I would hazard a guess that many of you didn’t. Regardless of your knowing-ness in the use of AI, it is all around us. And increasing in terms of its capabilities.
But with anything that is computer based, it’s bound to have issues. Bugs and glitches that need to be worked out. This is par for the course these days tho. Companies are always releasing products only to find out a few days or weeks later that something isn’t working, and then releasing an update or security patch.
But AI is no different. Well, maybe slightly different. As I mentioned, AI is used to identify information and then be able to present it to the user in some meaningful way. It can process large amounts of data in a short amount of time. But researchers are finding that if you add digital “noise” to an image, it becomes incredibly easy to trick AI.
So how could this become an issue? Well, what if AI was being used to do research. Let’s say reviewing legal documents for a law firm. Instead of paying someone to review the information, you could invest in some kind of AI software to do some of the work for you. I would like to point out that neither way is perfect. But we tend to assume that computer based systems can do so much more than humans, and I’m not sure that’s always the case. Getting back to my example though, AI in theory could process large amounts of data for a law firm.
But the challenge becomes when there are limitations to AI. Researchers took a passage describing the 1939 film Maisie. The first time around, the AI was 99.6% sure that the passage was about a film. Then one of the researchers misspelled one of the words, and the AI changed it’s analysis saying that it was 99% sure the passage was about companies. What is also interesting is that they changed the world “film” to “flim” in only one instance. It was in the passage in other cases, but still the AI thought about “companies”. Interestingly enough, that information wasn’t provided in the passage so where that information came from is a bit odd.
However, you can see how this could become an issue. But maybe not right now. These examples might be few and far between, but if left unchecked could become a huge issue in the future of AI. As I am writing this post, my mind goes to the movie Hidden Figures. Check it out if you haven’t seen it, but what I find interesting is that there is a correlation between humans and computers. Both how they are being used now, and how they were used in the 1960s. In the movie, we saw computers taking the place of people (who had the job title of Computer). But what happened was the machine computer wasn’t able to keep up to the person. In both examples, people program the computers or software as it might be today. So there is a large amount of human touch being applied.
Artificial intelligence is extremely fascinating. The idea that it can take data and process it in a way that makes our lives easier is incredible. It feels like the beginning when it comes to possibilities with this type of technology. I would love to see it being used to help people, but can also see it being used to possibly replace people, as it tried to do in the Hidden Figures example. One thing is clear though we will use this information to develop bigger and better systems.