Thread: ChatGPT
View Single Post
Old 02-09-2023, 03:10 PM   #46
Solecismic
Solecismic Software
 
Join Date: Oct 2000
Location: Canton, OH
I'm assuming that's an example of the computer programs ChatGPT can create?

I still don't understand the hype. Nothing's been written yet that passes the Turing Test. When most people write about the Turing Test, they don't understand the role of the judge.

The goal isn't to fool the judge for a minute or two. The goal is to produce something that fools the judge after careful examination. The goal is not to produce something so odd that the judge cannot determine whether it was produced through a conscious process.

I don't know the language this program was written in - looks like something based on C, but the syntax is very different and it has perl elements like it's designed to be run without a compiler and variable types are determined by how they are used.

But what's missing is context. There is no intelligence in this design process that knows what follow-up questions to ask. Even someone new to football wouldn't come up with a design like this. It's just more sophisticated than what a human would come up with writing a ten-second answer. Just like the text responses to knowledge-based questions.

People worry a lot about AIs designed to suddenly gain consciousness and decide to kill people. Seems silly to me. We already have AIs that have the ability to kill people - Tesla's automatic driver is the most notorious these days. It can kill if the human operating the vehicle uses it improperly (i.e., turns it on and falls asleep on the highway). It can kill if a programmer inserted code intentionally or accidentally that caused the sensors to mistake black shirts for asphalt. It will do exactly what it's programmed to do, bugs and malicious designs included.

Simulated consciousness is still quite limited by programming. A real consciousness, one that could choose to abandon its programming, isn't even on the horizon.

A sample sketch based on the above.

x = random(2);
if (x == 1) run the ball;
if (x == 2) pass the ball;

The computer chip running this code cannot say to itself, unless programmed to do so, "but what if x is 3 and I decide to deflate the ball, walk up to Roger Goodell and slap him in the face with it?" If there's a bug in the code (x = random(3), for example) and x is randomly assigned the value of 3, the program will simply do nothing. It won't think for itself, in some context, "huh, maybe I should punt?"

I think it's important not to lose sight of that. Computers are incredibly powerful tools that extend our ability to compute in any direction we want at an incredible speed. But they do not magically offer a new consciousness.

They will replace more and more jobs. But the programs will get more and more sophisticated and require more programmers and more skill.

Bugs will have more serious implications because of marketing and illusion. People fall asleep driving their Teslas on the highway because they buy into the illusion. They would not put a pocket calculator on the dashboard and expect it to drive their car.
Solecismic is offline   Reply With Quote