• Please review our updated Terms and Rules here

ChatGPT and code generation

Here's another one--a simple query and the answer is incredibly garbage.
Me: How many registers does the Control Data STAR-100 have?
Click to expand...

The answer:
The Control Data STAR 100 is a supercomputer developed by Control Data Corporation in the 1970s. It was based on a SIMD (Single Instruction Multiple Data) architecture and used a vector processor to perform operations on multiple data elements simultaneously.

The Control Data STAR 100 had 64 vector registers, each capable of holding 16 double-precision floating-point numbers, for a total of 1024 double-precision floating-point elements. The vector registers were named V0 to V63, and each register could be addressed by its number (0 to 63) or by its symbolic name (e.g., VR0 to VR63).

In addition to the vector registers, the Control Data STAR 100 also had scalar registers for storing individual data elements and control registers for managing the operation of the vector processor. The number of scalar and control registers varied depending on the specific model of the Control Data STAR 100.

Overall, the Control Data STAR 100 had a relatively large number of registers compared to other contemporary computers of its time, which contributed to its high performance in scientific and engineering applications.
Click to expand...

Wrong (256 64 bit general-purpose registers). Much of this seems to be excerpted from the Cray I description. The link must be "Control Data STAR-100 is a 70s supercomputer with vector instructions" and "The Cray I is a 70s supercomputer that has vector instructions"; therefore, the CDC STAR-100 is the same as the Cray I...
 
How about this one?

Me:
Describe the functioning of the Intel 8080 microprocessor using koans.
Click to expand...

ChatGPT:
In the realm of computing,
Intel 8080, a processor reigning
With its architecture simple and small,
Let's explore its working using koans, all!

In a world of zeroes and ones,
The 8080's journey begun
With 8-bit data bus and registers eight,
It processes instructions at an impressive rate.

Registers A and B,
Together they form a mighty team
Add, subtract, or multiply,
They perform arithmetic feats that will make you sigh.

The Program Counter (PC) points to the next byte,
The Instruction Register (IR) fetches it with might
The Address Bus, oh what a star,
It connects the processor to memory afar.

The Stack Pointer (SP) and the Stack,
Hold the return address and variables to pack
Push and pop, and jump to a sub-routine,
The 8080 handles it all with routine.

Flags in the processor,
Set and unset by each processor
Zero, Carry, Sign, and Parity,
Let you know if the operation's clarity.

In the 8080, there's no such thing as rest,
It's constantly churning, doing its best
To execute programs with speed and finesse,
In the computing world, it's surely a success.

And so we conclude this tale of the 8080,
A processor that paved the way, you know
For modern computing as we know it today,
Its legacy lives on in every way.
Click to expand...

Don't think it quite has the feeling for koans, but interesting nonetheless.
 
Given what it is (a slightly better parrot) it's not going to be difficult to find holes in its "knowledge." Except for finding really funny examples of bad output, I'm not sure what point you are trying to make anymore.
 
Go look at today's installment of eevBlog on YT. Dave's trying to make the point that it's a valuable program-writing tool--instead of a language tool. I think that's a mistake. Also note that at least one SF publication has halted submissions because it was getting too many ChatGPT-generated submissions. Or Vanderbilt U. sending out a consolation email about the Michigan mass shootings--generated by ChatGPT.

My point is that too many people are thinking that this thing is somehow "intelligent" and trying to use it in real-world applications where real human effort is called for. I can just imagine how many ChatGPT essays will be turned in by students as original work.

This can only end in disaster.

I trust that you understand now. :)
 
Oh, I'm aware of the shenanigans.

Those of us in the biz behind the scenes are asking why these things are being let loose. They are not ready for prime time; not even close. And we knew that a while ago.

The marketing types got too far in front of their skis. We're watching the face plant.
 
Another issue is validating the answer. You ask "who's the author of book x", it replies "John". Is it correct? Nobody knows, in order to verify it you need to do the effort you'd do if chatGPT didn't exist. I still don't see it capable of providing simple answers that you can trust, let along composing complex stuff.
 
You must log in or register to reply here.
Back
Top