3 ms·
In the late '80s I started studying AI in college, fully expecting it to boom. That did not really happen, until it did. Unfortunately, most of what I learned a
by kieckerjan 5y ago
In the late '80s I started studying AI in college, fully expecting it to boom. That did not really happen, until it did. Unfortunately, most of what I learned about AI back then is pretty useless now.
I started with symbolic AI: parsers, tree searches, expert systems, logic. Even then it felt like this approach was stuck in a rut. So I changed lanes and studied early ML: decision trees, genetic algo's, nearest neighbours. When I left college SVM's were all the rage.
All that time, neural networks were deemed interesting but unfeasible for anything realistic. Little did we know.
I feel that if I wanted to get back into this field, I might as well start all over again.
- zozbot234 5y agoSymbolic AI gets used quite a lot, we just don't call it AI. Even expert systems, often seen as a late-1980s dead end, are based on the same sort of technology one would use for efficiently querying any database.
- Smithalicious 5y agoIf the computer solves problems and we know what it's doing, that's software engineering If the computer solves problems and we don't know what it's doing, that's AI If the computer doesn't solve problems and we don't know what it's doing, that's Friday afternoon
- kieckerjan 5y agoGood one. I will remember that next time it's Friday.
- Smithalicious 5y agoIt would've been even better if I didn't butcher the formatting, but oh well
- disgruntledphd2 5y agoTo be fair, if you have the basic stats/code knowledge, you can totally pick up modern ML pretty quickly. You already know the difficult parts, the easy parts are pretty easy. This may not apply if you want to create new ML algorithms, but those jobs are rare in any case.