Posted on 07/05/2023 9:43:38 PM PDT by SeekAndFind
What are coal miners and journalists going to do?
I have a Computer Science degree.
I designed software.
I was taught how to design AI.
Coders are just secretaries that can type fast
that is about it.
They will be gone in a few years.
Who is going to verify that the AI-generated code will actually work?
When I saw this headline, it immediately made total intuitive sense. What better use for AI than to produce code? Whoever perfects this application will be instantly wealthy.
Google CoPilot. Nuff said.
what a load of crap
Same way they verified that the clot-shot actually 'works' -- unleash it on the public and wait and see what happens.
Glad I taught my kids auto repair, electrician skills, plumbing, and AC repair. I figured that they have long lives ahead and even though they did fine in college, things could change a lot...looks like that’s starting to happen.
Lol. You’re a bit too aggressive on your timeline. There are a few problems with this analysis. First, there’s a ton of existing code out there that needs to be maintained and improved. AIs are reasonably good at generating small chunks of code with specific purposes. Not so good at maintaining existing code or improving it. Further, before you can even get there on that, you need a better way of providing requirements to AIs. This is going to take time. A lot longer than 5 years.
“Who is going to verify that the AI-generated code will actually work?”
That is an interesting question.
How do we all know you will work correctly?
In the Mathematic world it is all Set theory.
How you choose to use that data is by design.
Computers by their design require 7 conditions for them to operate in a reliable predictable fashion.
AI is stuck with that.
People are not.
Y’all come up with a AI machine that thinks like a human then you have a BIG problem.
I can explain why 2+2 does not always equal 4.
I was a Math major at one time.
Much of what we do is based on beliefs.
Kind of like this sudden fear of AI.
Plus integrating with legacy systems.
I remember all of this talk when CASE tools came out in the 90s.
Well the real question is…do the H1Bs go home now?
We can always go back to growing cherries in the Santa Clara valley.
>> If AI fully takes over either programming or music
To be exploited by the untalented.
Y’all see how far CASE got.
Bottom line is the mathematics.
One of my specialties was Project Engineering.
This AI crap is just idiots that don’t know what the are talking about.
An example is the F35 they finally got the data set together
where the man in the cockpit has data he never had before.
I could program an F35 to independently kill only certain Russian aircraft, is that AI? No.
It's the same old snake oil that has been sold for the last 40 years, just in some shiny new packaging.
"Buy our Bla-Bla-Bal-omatic system and you can hire monkeys to make your applications. Monkeys are cheap and disposable. Our patented technology will generate all that nasty code for them, and it's going to be error-free. Call this number to learn more".
Coding can certainly be tedious work. It is also a minor aspect of software development. Modern programming languages and Integrated Development Environments (IDEs) drastically reduce the amount of coding required to make an application of specified complexity.
So, what happens? The customers require ever more complex applications, which require more coding to make them. And they require ever more capable IDEs to manage that code.
The critical factors are the design of a user interface suitable for the workload, the design of suitable back-end data storage and communications methods, and testable components with proper error detection and logging capabilities. Developers must be very adept at using complex IDEs and other toolchains.
It takes intelligent, skilled people to do that work. Most people cannot do that work. AI development is not currently directed toward doing that sort of work.
I have worked with AI-generated code from recent systems. It looks superficially pretty good. Then you realize that the software libraries referenced in the code do not exist or do not contain the methods called in the generated code. And the AI gets sulky when you ask for more details.
It is smoke and mirrors.
What an AI is really good at doing is finding component code samples posted by other developers, which do something close to what you may need to do in an application. It still takes a lot of discernment to select an appropriate sample and adapt it to what you need. But the AI can find good candidate samples faster that you can search the forums for the same stuff.
I presented ChatGPT with an engineering problem and asked for a solution. ChatGPT happily provided an answer. I asked ChatGPT to show its work. It provided the accompanying calculations. The answer was incorrect. I informed ChatGPT that it was incorrect and why. It apologized and revised the response to the correct answer. The bridge would have fallen down without human interaction.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.