Artificial Intelligence will fail us before it benefits us.
Follow the technology pundits and the startups in the world of Artificial Intelligence (AI) and you’d be left thinking miracles that will raise the human species are around the corner. Along with political and societal downfalls. Pick your preference, there’s a pundit for either or.
At some point, AI will become very useful; humans, industry and society will benefit greatly. More than likely anyway. But before we get to that point, AI will fail us. Not because the AI as a technology isn’t good enough, but for two primary reasons: humans and experimentation.
First, humans. We’re an odd bunch. We rarely use a technology only the way in which its inventor intended it. Such as Alexander Bell inventing the telephone to share opera music. We know how that turned out. When iron forgery was invented it replaced tools and weapons made from bronze. Humans made a lot more weapons and we promptly proceeded to bash each others heads in a whole lot more. Iron made weapons made the chariots of the time obsolete as iron tipped arrows and lances wrought havoc on them. Humans rarely think through the various implications of tools and technologies that we develop and throughout history, when one group of humans decide to just do nice, honourable things with their technologies, there’s always another group who don’t play by those rules. When it comes to AI, neither the Chinese nor Russian governments are debating the ethical uses of AI; they are quite happy to use it to control their populations and as weapons.
Secondly, experimentation. Right now, AI is in use in many places and ways that we don’t realize and it does particular jobs very well indeed. Like Spotify for your music and spotting cancers. In business, Quants use it for trading and make very nice profits. If you use any brand of smartphone, there is some form of AI at work. But for now, AI is really only good at very specific tasks. It takes an incredible amount of data and energy to make AI work. Humans are using their technologies and the companies that make AI, including the tech giants (Apple, Google, Amazon, Facebook) are experimenting with the data we nice humans give them doing the things we do such as buying and moving about the world. We’re all contributing to this experimentation phase of AI, helping companies make it better. Most people aren’t getting paid to train their algorithms and companies that buy AI products are paying the companies to experiment more so they can make their products better. It’s a bit of uneven playing field in many ways. And because it’s largely experimental, mistakes will be made. That is inevitable. The size and impact of those mistakes is impossible to predict.
For now and for some many years yet, AI will just get a whole lot better at doing little, specific things in very large ways, but AI is in its infancy. It is in an experimental phase. And humans aren’t entirely sure how to really understand and deal with it. AI may be the technology we have invented that is the biggest existential threat to humanity outside of climate change.
AI can help us with fixing climate change, improving our healthcare and more. It already is in some interesting ways. But we should be ever aware that humans do things in human ways. They cannot be predicted and we are still tribally divided as a species in many ways. We can expect AI to evolve very much along those lines. Go humans. So in the short term, AI is going to fail us quite a bit. Understanding this and preparing for it can at least help us mitigate the risks. Humans are a quirky bunch and so is how we use technology.
What do you think?
