Ai Dreams Forum
AI Dreams => General Chat => Topic started by: Zero on February 05, 2018, 09:26:08 am
-
I'd like your opinions about this.
Currently, I tend to think that speed of algorithms doesn't really matter: what matters is complexity (https://en.wikipedia.org/wiki/Computational_complexity).
My hypothese is that the obsession of speed is a cultural bias which comes from a distant time, when computers were, say, 1000x slower than they are now.
At that time, the evolution of computers was slower than today. Back then, devs had to produce programs that would run as fast as possible. A glorious dev was a fast dev. And there are young people today (remember the now gone "unreality" member), who were not even born at that time, but who are still influenced by that culture.
Now think about it. If I make a slow algorithm. I mean, for a good reason (for example, because I want to keep track of a lot of things). 5 years from now, it will really be fkin' fast.
I'm not saying that we should stop caring about speed. I'm just saying that, today we can afford huge "collateral" processing, even if it slows things down. Sure, some activities inherently need to be fast, like NNs.
Complexity is different. You don't want to freeze when you scale.
Where am I wrong? What do you think?
-
I agree, code can always be made to run faster, either by hardware upgrades, language changes or even just through profiling and code adaptations.
Initially you just need the code to run fast enough to prove your concept/ theory and get it working. The computer is a tool; it should help in your development not hinder it. You don’t want to be wasting time and resources optimizing code or learning a new language at the expense on concentrating on the main project goals.
The use of the computer should be as easy and seamless as possible to allow you use your mental resources to concentrate on the problems at hand.
:)
-
Zee,
An algorithm cut into brass gears survives millennia. Coming from a distant time in the past, around the year of the Consulship of Metellus and Afranius, ancient analog computers (Antikythera mechanism) were in use predicting the future.
+
+
+
+
-
Do you write sci-fi sometimes? If you do, I'd love to read it!
-
No, he just watches a lot of "Ancient Aliens" on the History Channel...as do I.
It might contain conjecture but it usually does prompt one to imagine the possibilities and open one's mind.
It's hard to argue with science.
==========================
Just as a note about the speed thing.
Remember when you played your first computer game like maybe Doom? It was difficult to beat the monsters and progress to the next level.
BUT...if you were to play that same game today, you'd go through it as if you had running at full speed slashing, hacking and shooting your way to an easy victory.
I think it might not be that we've already played the game but that we've become accustomed to faster computers and faster gameplay. Therefore making the old ones seem very slow by comparison.
I agree that speed is not always the measure of success for everything but it does say and do a great deal, for computing and problem-solving.