AI is Not Going To Kill Us All

Despite the oddly incompetent press put out by OpenAI (ChatGPT), and Anthropic (Claude), where they simultaneously tell us that AI will take all jobs, kill all humans, and also can you invest in my company, this is not a realistic possibility with current technology, or anything realistically being discussed.

Why? Let’s look at what AI actually is - an incredibly powerful set of technologies that do not actually think, they use statistics and probability to match content from your prompt with other content they have in memory, a document or somewhere else. There is no “thought” the way humans think. 

What keeps happening is a group of people who live in software code, and have only ever worked in a software company, think that software code and software developers are a good representation of other work, which is why they think that after AI has automated code, it’s just a matter of time before it automates other work, and we’re already seeing that this is simply not true. That same argument applies to prophecies of doom - AI is not the all-powerful “general intelligence” its makers seem to think. 

AI is still a product, and the makers of that product are responsible for its quality, something they’ve been reminded of pretty forcefully by the government (apart from some up for re-election politicians). These companies are now focused on ensuring that AI models they develop and release do what they’re asked, and only do what they’re asked, and don’t break things on the way. It turns out that the same quality issues that make AI generally useful are the same as those that make it safe - or at least are very similar.

This does not mean that there won’t be some high profile events, potentially very expensive events. There probably will. But it does mean that the news of our impending doom are exaggerated.