You’re right and I think the general concensus over the next few years will keep changing to support this idea. And probably go even further into vibe coding (to a certain degree) being a real thing actually companies do.
I don’t think a lot of people understand that even when the bubble eventually pops llms arent going away, and they legitimately are improving rapidly month over month with no sign that will abate. As they get better and better the cost to run models good enough for your specific task also lowers. I remember in 2022/2023 all the people here saying llms weren’t improving anymore and even getting worse, boy were they very very wrong.
A lot of people are going to be unemployed over the next dacade from this technology, a lot more people than just developers.
This of course is the fuckai sub so we will get downvoted for speaking the truth.
In case you missed the last memo, vibe coding is now more expensive than hiring a developer team, and the companies that hurried with layoffs are footing the bill.
That is to say, it was always the least cost efficient option, the AI peddlers just pulled people in with low prices until they seem irreplaceable.
Yes, but like I said in another comment models do keep getting better and better, and good models are getting cheaper and cheaper. As time goes on the price will continue to drop and capabilities will continue to grow. Frontier models will always be expensive, but a budget model that can do the same thing will come out 3 months after.
It’ll get cheaper bro, we just need more data, please bro, then we can use it to solve climate change, we just need more electricity, please bro, think of the capabilities bro…
If there is a future for LLMs, it will come after the bubble bursts and the small, specialized, locally hosted models take over, yes. I don’t expect them to be able to get good enough at coding to replace humans any time soon, though. You need logical reasoning and problem solving skills to produce code that’s of any value, and we aren’t getting there by text prediction.
[email protected]: I think the general concensus over the next few years will keep changing to support this idea. And probably go even further into vibe coding
Show us your vibe coded project. Oh, right, you won’t do it because you don’t want professionals to make fun of you. And you may also delete your comment like every other cog sucker.
I have never vibe coded, or hand coded anything. I took a beginner Python class but that’s the extent of my coding skills. My opinions are just from watching the last few years play out
¯(ツ)_/¯ I think as an impartial (nobody is actually impartial lol) observer I might have a more valid point than you, since I don’t really have anything to loose no matter which way the hat falls.
How can a project realistically tell if a contributor used an open-weight model and self hosted it without it being a de facto ban on AI for anyone not heavily involved in the project?
Open-weight models are still giving up control. For one, all LLM outputs are generally recognized to not be copy writable which undermines copy-left license. Additionally, unlike open source software it’s a lot harder to look at what represented in these models. A model can easily be designed to make specific plausible mistake as intentional backdoors, act worse on tasks relevant to an industry a country / maker cares about (like Fable has been designed), etc.
Very good points, the last part is especially interesting and probably something that will play out more over the years. Models with built in failsafes to sabotage work based on certain criteria, I imagine the usa/China have already started working on weaponizing this type of behavior.
You’re right and I think the general concensus over the next few years will keep changing to support this idea. And probably go even further into vibe coding (to a certain degree) being a real thing actually companies do.
I don’t think a lot of people understand that even when the bubble eventually pops llms arent going away, and they legitimately are improving rapidly month over month with no sign that will abate. As they get better and better the cost to run models good enough for your specific task also lowers. I remember in 2022/2023 all the people here saying llms weren’t improving anymore and even getting worse, boy were they very very wrong.
A lot of people are going to be unemployed over the next dacade from this technology, a lot more people than just developers.
This of course is the fuckai sub so we will get downvoted for speaking the truth.
In case you missed the last memo, vibe coding is now more expensive than hiring a developer team, and the companies that hurried with layoffs are footing the bill.
That is to say, it was always the least cost efficient option, the AI peddlers just pulled people in with low prices until they seem irreplaceable.
Yes, but like I said in another comment models do keep getting better and better, and good models are getting cheaper and cheaper. As time goes on the price will continue to drop and capabilities will continue to grow. Frontier models will always be expensive, but a budget model that can do the same thing will come out 3 months after.
It’ll get cheaper bro, we just need more data, please bro, then we can use it to solve climate change, we just need more electricity, please bro, think of the capabilities bro…
Just 75 more gas turbines killing local residents and I swear grok will be good!
Now you’re getting it
If there is a future for LLMs, it will come after the bubble bursts and the small, specialized, locally hosted models take over, yes. I don’t expect them to be able to get good enough at coding to replace humans any time soon, though. You need logical reasoning and problem solving skills to produce code that’s of any value, and we aren’t getting there by text prediction.
I generally agree! Local self hosted models with smart and competent users capable of properly using them would be the ideal future.
Show us your vibe coded project. Oh, right, you won’t do it because you don’t want professionals to make fun of you. And you may also delete your comment like every other cog sucker.
Have a nice day anyway.
I have never vibe coded, or hand coded anything. I took a beginner Python class but that’s the extent of my coding skills. My opinions are just from watching the last few years play out
Is this a fucking joke?
It could be, are you upset or enjoying the convo?
Baffled by your ability to contine breathing seeing how braindead you are.
So you have no clue what the hell you’re talking about then. Cool thanks for confirming we are all correct in ignoring what you’re saying.
¯(ツ)_/¯ I think as an impartial (nobody is actually impartial lol) observer I might have a more valid point than you, since I don’t really have anything to loose no matter which way the hat falls.
if you believe that, you have already lost
So why help by giving training data and give a unnerving lever of control of open source projects to a handful of companies?
Use open weight Chinese models that don’t collect your data.
How can a project realistically tell if a contributor used an open-weight model and self hosted it without it being a de facto ban on AI for anyone not heavily involved in the project?
Open-weight models are still giving up control. For one, all LLM outputs are generally recognized to not be copy writable which undermines copy-left license. Additionally, unlike open source software it’s a lot harder to look at what represented in these models. A model can easily be designed to make specific plausible mistake as intentional backdoors, act worse on tasks relevant to an industry a country / maker cares about (like Fable has been designed), etc.
Very good points, the last part is especially interesting and probably something that will play out more over the years. Models with built in failsafes to sabotage work based on certain criteria, I imagine the usa/China have already started working on weaponizing this type of behavior.