Why we don’t use AI, at least not yet

I can’t open up any social media or news app without seeing something about AI these days. Usually what I see is something about how AI is going to take all our jobs or about how much Nvidia is worth. Neither of these strike me as particularly good news, and one of them is totally false (can you guess which one?). We often have discussions here at Ethos about AI and when people find out what I do they always want to ask about AI and how we use it at Ethos. Well, we don’t use it. At all. At least not in the product and not right now.

AI is one possible solution to a particular problem. Usually there are several ways to solve that problem and AI is only one of them. For example, most of the AI that is being used by people today is used as a replacement for search, but the only reason that they need a replacement for search is because search is terrible and has been terrible for years. Now one possible solution to fixing search is that you just bypass fixing it altogether and create a totally different way to search for something, like using a chat interface. You can start to see our logic here, the root problem is that search is terrible. Rather than using AI at all you could just fix the root problem.

Most of the problems we are trying to solve at Ethos have multiple possible solutions, problems like updating a donors information in a way that makes sense, or attributing a donation to the right person and getting them a receipt that looks great, or running a report about your organization’s transactions in a way that anyone can do it. All of these problems have multiple solutions, solutions like a better designed user experience than we have seen in the past, or specially designed interfaces that make something dead simple to do. Could we use AI to solve these problems? Of course we could, but in most cases AI is the most half-baked, resource intensive way to solve them. This is all leading to the main reason we don’t use it, AI makes mistakes.

The problems we are trying to solve are important, they involve money, they involve a non-profit organization that works on real world issues that people are passionate about, and they involve people’s personal data. AI is one possible solution but AI is also not a finished product in and of itself. Anyone that has used AI for something complicated or important can tell you that AI makes mistakes, lots of them. These mistakes are fine when you are generating funny images for your friends, but if they were involved in something involving people’s money, or their personal data, the results could be much worse. So yeah, we don’t use AI because AI isn’t finished yet. We’re not saying we never will, but right now at Ethos we think our time is better spent trying to solve the root problem.