- Computers do a lot of amazing, seemingly magical things because:
- There's a lot of layers of abstraction hiding many complex systems and components, and it's nearly impossible to understand it all from end to end
but...
- Tech industry hucksters love this characterization because it makes their products appear flawless, or capable of much more than they really are
- If pressed on a point, they can defer to some highly technical publication, even if they've badly misrepresented it, or if the publication was fundamentally flawed
- But that's enough to get people off their back, because most people can't wade through the overly-complex specifics to understand if any of it makes sense or applies
and so...
- In this situation, it's important to remember that a computer is a tool that does a little math to make signals
- Fundamentally, we can't expect them to perform tasks with efficacy if we have not fully understood the problem and how to come to a conclusion about the problem
- Machine learning and AI products are frequently marketed using flashy magical marketing
- They're often just very large statistical models, and as such, we should scrutinize AI as we would statistics
- This means we can't expect a computer to solve a problem that we haven't fully understood
- They can perform a lot of calculations very fast
- But, remember, they're still just tools. We can't expect a computer to perform something like legal and ethical judgments using a bunch of statistics in the same way we can't expect a human to write a formula for such judgements
- This is especially true if we don't fully understand the model we used to compute these probabilities
So if someone is trying to sell you something magical (read: AI), take a big grain of salt and hire some trustworthy people to understand all the technical and ethical consequences of the system
