4 ms·

People often bring up the quip that "as soon as something is AI, people stop calling it AI" but I think pretty much the opposite has happened, the term has been

People often bring up the quip that "as soon as something is AI, people stop calling it AI" but I think pretty much the opposite has happened, the term has been expanded so much that almost any stochastic computation that solves some computational problem is now shoved under that umbrella. However unlike jordan in the article who takes that as an opportunity to focus even more on engineering and practical applications I think it's well worth going back to the original dream of the 1950s. I'd be happy if people return to the drawing board to figure out how to build real, artificial life and how to create a mind.

tluyben2 5y ago
People in my circles stopped saying AI and switched to ML after it 'did something previously identified as more or less intelligent'. I see this is more or less the norm. That makes some sense because once we understand how things work they become repeatable and thus not-that-intelligent in the eye of many. And that, again, makes sense considering we are nowhere near anything that has the 'I' in it: it's pattern matching which is statistics and it works on a massive scale due to tech/gpu and ago advances but there is no 'I' there. The most basic mistakes are not correctable by the most advanced stuff we have and as such, it gets things it is 'bred for' wrong all the time, with sometimes disastrous outcomes. This stuff should be in labs, not identifying gun shots, faces of criminals, driving cars, identifying 'fraudulent' people and such. Because of the 'I' missing, this is a nightmare and it will get worse.
darksaints 5y ago
The description of the Turing test certainly alludes to a test of machine intelligence. So to some degree it is appropriate for AI to have a large umbrella. My problem with the endless battle to define AI is that there are a hopelessly clueless cohort of people with money that chase the term like its their golden goose, and therefore anybody that wants funding needs to market their thing as AI. And I think a lot of researchers cringe at the wanton use of AI because it devalues the work that they're doing. I just give it the shoulder shrug of approval - "I get it, you need funding too". And from that perspective I really wish that tree-based ML methods and logic programming languages and rule engines were still called AI, because they're really cool but horribly neglected because theyre not the latest thing.

tluyben2 5y ago

People in my circles stopped saying AI and switched to ML after it 'did something previously identified as more or less intelligent'. I see this is more or less the norm. That makes some sense because once we understand how things work they become repeatable and thus not-that-intelligent in the eye of many. And that, again, makes sense considering we are nowhere near anything that has the 'I' in it: it's pattern matching which is statistics and it works on a massive scale due to tech/gpu and ago advances but there is no 'I' there. The most basic mistakes are not correctable by the most advanced stuff we have and as such, it gets things it is 'bred for' wrong all the time, with sometimes disastrous outcomes. This stuff should be in labs, not identifying gun shots, faces of criminals, driving cars, identifying 'fraudulent' people and such. Because of the 'I' missing, this is a nightmare and it will get worse.

darksaints 5y ago

The description of the Turing test certainly alludes to a test of machine intelligence. So to some degree it is appropriate for AI to have a large umbrella. My problem with the endless battle to define AI is that there are a hopelessly clueless cohort of people with money that chase the term like its their golden goose, and therefore anybody that wants funding needs to market their thing as AI. And I think a lot of researchers cringe at the wanton use of AI because it devalues the work that they're doing. I just give it the shoulder shrug of approval - "I get it, you need funding too". And from that perspective I really wish that tree-based ML methods and logic programming languages and rule engines were still called AI, because they're really cool but horribly neglected because theyre not the latest thing.