From the findings
There is a popular website most likely used and not named in this study called Topcoder where the competitive challenges were pulled from.
Community - Competitive Programming
As someone who has actually used the website and done the challenges, when they say the simplest problems it might mean something like.
An actual problem from the site (you need to register to see it)
Topcoder Arena
They also give you the function definition and parameters too so you already are given what structure your function should be constructed
This is a simple "if then else" problem. Where you're doing a conditional check on if something is true, in this case the string conditions are met for the strings where its a certain length, and the characters can only be A and B and there must be K amount of pairs in the string to satisfy a true condition. If found, return the string, else return "".
Fairly simple problem and this complicated AI programming could only do problems of this difficulty. This is like 3rd week stuff you do in your intro to programming course freshman year, first semester in your 4 year CS coursework.
In short, it will be A LONG TIME before automation replaces programmers. Also to iterate Mark Cuban sounds like an idiot. Not saying it won't ever happen but all this recent fear mongering of the idea in the past year or two is weird given the technology isn't anywhere near being there. Why now? Can't help but feel its part of a propaganda campaign to get workers ready for more unemployment and become comfortable with "Basic universal income".
I never disagreed that the stuff it could do now is reallllly simple stuff. but lets not forget. a lot of our jobs are really simple if you break it down to if, then statements.
no matter how complicated you THINK the programming is. I can break it down into if, then statements, nested within themselves.
The reason this is the case is because the output you want is very simple.
For instance developing an animation based on real life physics. is difficult for many computers to handle because real life physics has a ton of equations going on at once and a lot of factors that can change the thing you want to animate. which is why games rarely use "real life" physics. but if you break it down piece by piece, you can come up with if then statements for each thing in order to make said object move from point A to point B.
because the question is usually simple.
I want my google maps to give me the best (least amount of traffic, shortest distance traveled) directions from my job to my house at 4pm.
well, to us non programmers it looks all sorts of complicated. but if you start to break apart the logic. it isnt. it just takes a lot of programming to get something so simple done correctly.
how far is point A from point B? How many ways can one person get from point A to Point B? what is the top 3 shortest routes?
If person A is trying to get from point A to point b. Bring up 3 shortest routes.
How much traffic is usually on the freeway's this time of day(database full of info showing how long it takes cars to drive X amt of miles on a thursday at 4pm)?
do the math, if best route 1 is the fwy 100% of the time. yet that route is now congested and we foresee it remaining that way based on actual live sensors and previous historical data. THEN, move to Shortest Quickest Route #2...
nest that statement. if route #2= too long(or false), then Route #3. nest that statement.
it would look insane. but thats the point. you code that in a particular language it may not look like a bunch of if thens. but thats actually what it is.
What Ai some what is versus what the super AI could be is two different things.
There is some things they call AI, where people program the intelligence into it. and it takes that programming of patterns and uses its ability to now capture all of the most likely scenarios and produces the results that you humans would normally produced based on X number of factors for these scenarios.
If i'm the Ai and i have been programmed to know, on thursdays you usually go home around 4pm. I already know i should send you the google map showing your travel time from work to home. Sure i could be off a bit if you decide to stop off at your brothers house before you head home. but if you keep stopping off at your brothers house once per month. I can learn that pattern. and realize, approx once per month around week 3 of the month usually on the 3rd Thursday, this guy stops at his bros house before heading home. so i will send him directions "hey do you want directions & traffic to you bros house?" or are you heading home ?
That type of stuff is already occurring. are real humans still feeding it data?l Yep. are humans going thru the newer data and making sure the AI reads it correctly? yep.
this is why it isnt complete yet where it can do it all on its own.
But the super AI you guys are referring to where it will be able to think on the fly like a human and even come up with its own ideas, thoughts. is very far away...I THINK. The reason i say i think is obvious, like you said their still on the basic stuff. BUT, computer/tech/science in this computer era has taken leaps in a lot of fields, that in the past would've taken 100 years to reach, because the better computers get the bigger the jump in ability they have. quantum computing is here, but not hear for us home consumers. .....YET.