AI Fails
- Thread starter ISULibrarian
- Start date
Grabbed this screenshot from my Google newsfeed on Friday morning, more than 24 hours *before* the actual game 
Not a "fail" per se as it did actually work, but more of a concern. I've been playing with Python (via VS Code and Jupyter notebook) as a lot of people at work are using it for data analysis automation. My concern is people are using AI for code generation when they don't actually have any expertise in the field. In my case, that's reliability engineering.
I was working on redoing a Bayesian analysis I'd already coded up in Matlab to make it more accessible for other people. As I was typing, the AI engine automatically wrote the complete log-likelihood formula, presumably based on prior context and a variable name. Now the formula was right, which was good, but it's not something that someone outside reliability is going to recognize as right or wrong and there was no reference to where it pulled the formula. If AI had hallucinated the formula, or given the version for a different type of data set or a different distribution, how is a non-reliability person going to realize that? These analyses are used to make multi-million-dollar decisions on warranty policies, proactive repair campaigns and recalls. I'm afraid it's going to reduce critical thinking and the perceived value of expertise.
I was working on redoing a Bayesian analysis I'd already coded up in Matlab to make it more accessible for other people. As I was typing, the AI engine automatically wrote the complete log-likelihood formula, presumably based on prior context and a variable name. Now the formula was right, which was good, but it's not something that someone outside reliability is going to recognize as right or wrong and there was no reference to where it pulled the formula. If AI had hallucinated the formula, or given the version for a different type of data set or a different distribution, how is a non-reliability person going to realize that? These analyses are used to make multi-million-dollar decisions on warranty policies, proactive repair campaigns and recalls. I'm afraid it's going to reduce critical thinking and the perceived value of expertise.
Presumably it will get better, but I agree. Right now it's an undeniable productivity enhancer, but the problem is if you don't have a certain level of expertise, you won't recognize when it's wrong. (When applying in certain fields.) People without the expertise and using AI won't catch that though.
Sometimes it's just obviously wrong though.
Cyclonepride
Thought Police
I've only used AI a couple times for very rudimentary things (Facebook banners) and AI has one hell of a time spelling words (that you give it directly) correctly. No idea why it can't handle such a basic thing like that.
