http://nautil.us/blog/heres-how-well-kno...-conscious
So, I'm not super-knowledgeable when it comes to computers, but I do understand the mathematical side of things like data mining, machine learning, and statistical algorithms (probably better than the vast majority of science and tech writers) and I thought I'd weigh in on this...
I think that lot of people (even some people like Elon Musk and Bill Gates) fail to understand the difference between
algorithms and actual
thinking. An elementary school kid that's learning to do multiplication and division on paper almost never really understands why the step-by-step procedure they are taught to do that works, they just learn to do it correctly to get the right answer. That's basically what an algorithm. It's fine for them because for something like arithmetic it will always give you the right answer if you do it right.
That's essentially what an algorithm is and what any computer program is. It's a set of instructions involving performing logic and arithmetic operations on some data that it is followed to get a result. The instructions are followed to the letter without exception even if contexts arise where the process would have to altered for some particular reason the program didn't see. What people refer to as AI is really statistical algorithms finding complex patterns in data using math the vast majority of people don't even realize exists. The only difference the programs that people call AI computers from 50 years ago is that the former involves more sophisticated procedures that require better hardware.
In my mind, the way to know that we've achieved
true AI (that's what the article is about), a computer has to be able to comprehend plain English. That doesn't mean Google or Alex being able to answer how many ounces are in a cup. Dogs and birds can learn to recognize what words and short sentences mean.
As an example, an MIT computer scientist created a computer program that could assign grades to student essays that corresponded very closely to grades given by professors. Of course it cant actually read and comprehend, the program looks at thing like average sentence length, the number of unique words used, and the frequency of certain letters. That's not understanding english, that's a calculator classifying papers based on patterns a person could never recognize. A person could write grammatically-correct gibberish and get an A from the program.
It's the reason that MS Word can't fix every grammatical error and why email software will put important emails in the spam box. It's why Alexa and Google home can't answer questions that aren't more than several words and it's the reason why Google can answer "What is 4258x3561" but is useless if you put 5th-grade word problems into it. It doesn't comprehend text in any meaningful way, and I don't really see us getting any closer to that.