Not developing AI. We've been working on and with ANI for a very long time, and constant research on what might make AGI possible. What side railed everything is some companies found out they can do lots of amazing things with LLMs, a subset of AI science, and investors will pay them to make better versions. They talk a lot about how what they're doing will lead to AGI, but these people (or the scientists doing the actual work) know better than that. LLMs can possibly be part of what AGI uses, but it's on the wrong branch. And they've given up following the other branches because this one is paying off.

So in short, we could have done better (both ethically and core results), but once again money got in the way.