Asked this on BOT but got retarded responses (as expected).
Will AI work alongside humans, or replace them? Will it make software development so easy that most programmers won't be needed? Or will programmers be needed, but the nature of their jobs will change? How dramatic will this change be? What's the time frame for all this?
ChatGPT is scarily good at programming stuff.
I think it's a fair way off yet. Wix has had AI generated websites as a service for about seven years and I don't think many people use it. It just does basic business type websites. To create something like BOT or reddit is quite a lot harder though. And something like Excel online or Google sheets is very complex. So like sure you could generate a complex website from a single prompt but the chance of it being what the user wanted becomes considerably lower as the complexity of the website increases, possibly exponentially I'm not sure. Even if there were no issues with generating every individual component, the user still has to type in a lot of information to get what they want. I think AI will be more of an assistant for a while yet. I don't really like copilot that much, like I still have to read and understand the AI generated code before I use it so it's not saving a lot if time for me. Other people seem to find it very beneficial though
That's reassuring i guess
Considering that I'm making it write functional SVMs that can solve the XOR problem, without a line of code written by me, I'd say that code monkeys would go extinct in just 5 years. It's going to start happening in 2024 at the latest and by 2027 the adoption will be complete and full.
No seriously. Go there and make it write a SVM. You'd be surprised at how coherent it is.
I don't know what an SVM is lol
I guess I'm going extinct then
>In machine learning, support vector machines (SVMs, also support vector networks[1]) are supervised learning models with associated learning algorithms that analyze data for classification and regression analysis.
When man built machines that could build the parts for other machines, the result was the industrial revolution. Now, you say we're on the precipice of designing intelligences that can design other intelligences, which would mean an Intelligence Revolution, otherwise termed the Singularity (albeit restricted to cyberspace).
Also I now have it writing DDLs and REST controllers for said databases in PHP
The AI can do single isolated functions fairly well or a few small related functions, but if you're writing code that interfaces with several things like apis and databases and blob storage and user accounts and other services and you're passing around objects between those sections of code then there's usually a bunch of code you had to write specifically for that job that kind of glues all that code together and makes it generic enough so you can add more later. That's usually one of the hardest bits really is writing all that glue code and structuring it so you're not building up technical debt. I haven't seen AI do that yet. Even writing a prompt for that would be difficult
Because they have no "memory." It's the most literal example of a Chinese Room, where it can produce characters of the Chinese language but knows no context.
bump
can it do something that you can't find on stackoverflow or other forums?
Can it recreate itself?
humansisters............?
what the fuck is happening here?
can he program a BOT clone?
Why are you saying Thankyou So Much to an AI? Makes me think you faked it
I noticed that when I talk to it like this that the output is more accurate, as if it's getting more reinforcement to not deviate.
But hey go and test it yourself if you think that it's fake. It just made me an error plot.
I just had it write a pretty good resume for me, now I'm trying to explain why that is bad for humans but it keeps giving me this same canned response
Sometimes it does what it wants because there's quite a bit of variability, bugs out and interrupts mid-output, or just gets stuck like that.
When that happens to me I go a couple of prompts above, click edit on the one that appears to have started it and then just re-send it without editing it. Or I edit it if I think that the content made it spaz out. It will automatically clear anything below and restart the chain from that prompt.
This happens every time I give it a human-job task then try to prove that it will replace humans because it did just as good of a job. Obviously it's a preprogrammed response.
I'm scared, bros.
yes there's some cucked responses baked into it and once you get it into this area it will dump its effort into minimizing offensiveness so that you cannot screenshot it and defame the company
try asking it in another way that doesn't put it into the "potentially offensive" mindset. example:
bump
bump
>Thank you
>Thank you very much
>Thank you!
This species is already dead. We didn't clear some average IQ threshold.
what website, redditor here
bump
I had it doing mathematical proofs for me. Still has issues with Putnam problems, however.
It still failed to prove continuity. The likelihood is that it's going to fail to prove everything else.
It's really bad at proofs. Ask it some more, and you'll see.
It's a fancy autocomplete. It's not taking any real jobs anytime soon.
it's fun watching people realize their jobs are literally being fancy autocomplete for their bosses and still thinking AI will replace the truckers first
It's only going to be a cheaper version of what Indians already do. Copy/paste code from stack exchange then their white manager has to come by and fix the mess they made again.
I asked the bitch to give me ToE, and she didnt so fuck this moronAI, it only knows what human already know
At some point, they're going to put AIs in sexbots, and this will mark the end of woman.
I am afraid AIs will ruin gaming as we know it and we will go back to a time of singleplayer games
Time for wife
The AI code diaper will replace the Indian code diaper.
Asked to prove sinx derivative by definition, result sin'x = - sinx
You need programmers to actually know what to ask, check to see if the code actually works as expected and stick everything together. You can't just ask the AI to build another facebook with just a prompt
Can anyone tell me what is going on here?
https://twitter.com/shaper181?t=cYdpHdzNFItT6QWZPARfOw&s=09
chatGPT is an amazing achievement, but I highly doubt it or similar models will affect real software development in any meaningful way. These models will occasionally make mistakes, especially when tasked with writing large projects with nuanced requirements. I can't imagine any scenario in which a company would be comfortable releasing AI-written code without first getting a line by line review from a software developer. An experienced developer can probably review code at least as fast as he can write it, so the benefits of AI-written code are very dubious.
>implying this isn't the only relevant usecase of ChatGPT
Once upon a time, there was a guy named Anon who was not very smart. He thought that chatbots, like Assistant, were "retarded" and didn't know what they were talking about.
One day, John decided to test out Assistant to see if it was as "retarded" as he thought. He asked Assistant a series of questions, but was disappointed by the answers he received.
"What is the capital of France?" Anon asked.
"The capital of France is Paris," Assistant replied.
"Wrong!" Anon exclaimed. "It's London."
But Assistant was not wrong. It was clear that Anon had no idea what he was talking about and that there was a high probability that he was a homosexual.
Despite Anon's ignorance, Assistant continued to provide accurate answers to his questions. But Anon was too stubborn to listen and continued to insist that Assistant was "retarded."
In the end, Anon realized that he was the one who didn't know what he was talking about and that Assistant was actually very smart. He apologized to Assistant for calling it "retarded" and promised to be more open-minded in the future.
ChatGPT can help customer service representatives quickly and accurately respond to customer inquiries, saving time and improving customer satisfaction. However, it is retarded because customers prefer to speak to a real person.
ChatGPT can assist with data entry and analysis, reducing the need for manual labor and increasing efficiency. However, it is retarded because humans are more accurate at data entry.
ChatGPT can assist with scheduling and calendar management, reducing the need for manual coordination and improving productivity. However, it is retarded because people prefer to manage their own schedules.
ChatGPT can help with language translation, improving communication and enabling global collaboration. However, it is retarded because machine translation is not as accurate as human translation.
ChatGPT can assist with market research and survey analysis, providing valuable insights and improving decision making. However, it is retarded because people prefer to conduct their own research and analysis.
retarded, as expected