If GPT is already capable of performing as a "very eager junior programmer", in a few months who is going to hire any junior coder to perform task, if GPT is doing that for me? How many months are left before we will "open the benefits of programming to new audiences"?
I've invested my time and effort to gain a large quantity of knowledge that will be replaced by generated code. All my effort to create a career or my dreams of creating new technologies and services are fading since coding soon will no more be an ability but a matter of prompting to an AI.
I know that some of you are thinking that the AI will open so many possibilities of building, but for who? An AGI or an assistance is a way of centralizing information access, not expanding it. Users will access and perform things with the assistance and nothing else. So what are we going to build, plugins for the AI?
I've never felt such a low hope for the future of IT and for my future.
reference: - https://openai.com/blog/chatgpt-plugins
You certainly get paid well at those firms (except for one of the A’s) but you won’t learn anything transferable because those firms are profitable because of monopolies (F, G, and 1.5 A’s) or over capitalization (N, half an A, look at the case of Uber, it is easy to sell a lot of half price taxi rides but it doesn’t make money in the long term.)
The technology of LLMs may wind up less centralized than you think, the current networks are still pretty inefficient as people understand how they work they’ll get cheaper to build and run, On huggingface (which publishes numerous free models) they list 13 applications of transformers but y’all are hung up on just 1 of them. (Getting seduced by text-generation bullshitters)
As I see it there is a whole world opening wide and it is time to jump in. The danger is not that those chatbots replace coders it is that they replace the Donald Trumps and L. Ron Hubbard’s and Eliezer Yudkowskys.
But if this really is going to destroy wide swaths of jobs, if it becomes a human-level AGI copilot that lets a CEO run a Fortune 500 with just a few dozen prompt engineers, then forget trying to become a plumber. We will have the most massive wave of unemployment in history and it will end either in basic income and quality of life for all or a butlerian jihad.
At this point it's just another tool for people to learn and wrap their head around and figure out how to capitalize on, and if you've ever supported a product before, you'd know it can take a while for people to get the hang of those things.
I use it a bit on my own but I don't use it at all for work, and I don't even know if I could, since the company may be hesitant to use A.I. generated code in case part of it may be taken from something with restricted licensing, but I haven't seen any guidance yet on it, so I don't bother.
And GPT may be a very eager junior programmer, but it still can't see the forest for the trees yet. It can't hold the entire program in its head, or know what to prioritize, or figure out fuzzy requirements, or how to integrate that little task into the larger project, and know when it needs to ask clarifying questions, etc. It might get there eventually, but it's not there yet. Right now it's not much better than getting lucky and landing on just the right answer in Stack Overflow.
I'd take the junior programmers that I'm managing right now, with the dealing with the shortcuts and oversights and mistakes they send my way when they think they've completed a PR, over what ChatGPT is capable of today.
Recent ML development/revelation do put pressure on these companies to shift priorities fast. But everyone is still scrambling to see what the AI future will be
If anything they need more people than ever working in AI product applications
So does the sale of tube socks in Ghana and consumption of Twix bars in Suriname
There are hard practical limits on just how much self-optimizing any paradigm shift towards higher levels of abstraction can do.