Getting a Grip on ChatGPT

“Any sufficiently advanced technology is indistinguishable from magic.” — Arthur C. Clarke

There’s something strangely unbalanced about the human reaction to a lot of new things, particularly things that the average person doesn’t fully understand. We tend to either over-react or under-react, seldom hitting just the right response. The recent wave of over-reaction regarding OpenAI’s ChatGPT tool seems to be a great example. It’s the end of the education system. The internet is going to break. AI tools like ChatGPT are going to learn how to program, take over everything and rule the world. Apparently they’re acting in threatening and “dark” ways, etc.

This article is a wonderful piece which helps, I think, put it all into a little perspective, and, boy, perspective is sorely required at the moment.

Perhaps the most important message in it is one single line… stop anthropomorphising AI. Seriously. Every time we start viewing and thinking about AI in anthropomorphic terms, we immediately start to imbue it with abilities and powers it just does not have and the world becomes a lot scarier and freakier. We allow every fanciful SciFi movie and book to let fiction and imagination slant our view of what is actually real. Then we over or under-react.

AI tools are absolutely amazing, and ChatGPT (along with other large language models) has enormous potential to massively improve the way humans access, understand, synthesise and utilise information from an enormous range of sources. It’s a very cool piece of technology. But, it is not sentient. It has no capacity to be dark and dangerous in any kind of personal way. It is a massively sophisticated search engine, one which sometimes comes up with the wrong answers. It is useful and fascinating, and the opportunities it offers up are quite transforming but it is just a tool.

Like using any kind of internet search, its utility is heavily reliant on the person wielding the tool. Ask the right questions in the right ways and you can get some amazing results. Go down certain types of interrogative rabbit holes and you’ll get weird and strange responses back. This is not because there’s some kind of demonic sentient beast lurking in the bowels of this tool. It’s just limitations in the tool’s ability to properly deal with strange lines of questions. Old school systems would just bail out and tell you nothing can be found, try again. A true human would probably tell you that you’re being stupid or rude. ChatGPT just robotically keeps on trying to give you back something, even it becomes more and more obscure and nonsensical.

This new type of tool is amazing and it’s going to take us a little while to find our footing and figure how it truly helps and where it hinders. But, like all of the cool tools that came before it, we’ll probably look back at some of the reaction and wonder what the hell people were thinking when they freaked out back in the old days.