Sure, our models have close to no intelligence. They're just automating what we've trained them to do. But that power makes them extremely useful in the real world. Just not a panacea.
Show this thread
Dismissing machine learning because it can't make sense of what it *hasn't* seen before it quite short-sighted. It is immensely valuable to be able to automatically recognize *what you are able to label* -- especially on hard pattern recognition problems, at super-human accuracy.
-
-
-
I see too many people taking the stance "ML does not magically solve every problem, therefore we shouldn't use it". Instead, engineers in every industry should be asking, "what can I do with ML?", while keeping in mind its limitations. Turns out, there's *a lot* you can do.
Show this thread
End of conversation
New conversation -
-
-
I feel like we are on the verge of a golden age of math and science as a result of this:https://twitter.com/nik_hultgren/status/1218981519319977986 …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Sure it's useful. But you should remember that all is reactionary as we have seen by ML researchers that "AI is electricity" and "DL is Software 2.0" claims from **far-sighted** ML researchers.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Even here super-human accuracy is an ill-defined term. ML means/promotes mediocracy. Yeah, that has some uses for sure.
-
I literally work on replacing human led audits in retail using relatively simple neural networks. The only reason we survive is because our system is significantly more accurate than humans at the same task (and because we can do it about 21x more often).
End of conversation
New conversation -
-
-
My problem isnt with ML it is with the lack of sufficient GI models that people who use ML have. when you model GI you'll see ML currently doesnt build 4d memory schema that mirror our own. It doesn't attend, observe, memorise, generalise, bootstrap, recall like we do.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Agreed! It's pretty much the same insanity as saying "we shouldn't build any machine because they are only capable of doing what they were built for..."
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
You forget everything that is unsupervised from embedding to anomaly detection. We can measure and evaluate what we could not before.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Agree. Automation has always been met with resistance through out centuries as we moved step by step in mechanization of tasks. ML is unique because as we are exposed to large quantities of information - which is what the case is...
-
... we can’t make sense of it beyond contain point. Recognizing patterns in large quantities of data is absolutely beyond human capabilities....
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.