[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: images (2).jpg (27 KB, 631x486)
27 KB
27 KB JPG
its over
>>
File: big.png (204 KB, 2601x1286)
204 KB
204 KB PNG
>>108348635
>>
>>108348635
>Line go up
I have no idea what this means.
>>
>>108348646
Obsolete retard #1 - confused parameters for capability. Replaced and destitute.

>>108348651
Obsolete retard #2 - can't read. Replaced and bankrupt
>>
>>108348635
What is this terribly made graph supposed to mean?
>>
>>108348651
As the years go by the models are getting worse/slower. GPT-2 accomplished tasks in less than 30 minutes while newer/bigger models like Claude Opus 4.6 take 15 hours to do the same thing. How embarrassing!
>>
We are so fucking doomed. Look at how stupid and confused we are
>>
>>108348635
You're comparing a general chat model to a model specifically trained in coding. Try the same thing with 5.2 codex max
>>
>>108348635
I never understood this diagram. What does it mean that it now takes 15x times than 5 years ago?
>>
>>108348684
Lmao best comment
>>
>>108348684
It is quite a powerful graph, LLMs have taken over.
Truly the end of humanity.
>>
>>108348699
AI salesmen allege that their AI can do a task that takes a human roughly N hours 50% of the time.
>>
>>108348651
>>108348668
>>108348699

Please for the love of god, please im begging you on my knees to tell me that you aren't employed as software engineers, or even anything that revolves around operating a computer
>>
>>108348695
This is our one hope, that the evil jews saw profit in our suffering and targeted agentic coding somewhere in 2024/2025, and that this is going to s curve out soon
>>
>>108348723
I am not employed as a software engineer, however I do work with computers.
Sorry anon. But hey maybe you can explain to me what the confusing graph is about and that way we can fix the situation.
>>
>>108348723
The diagram is not clear out of context like this. What does it supposed to represent the hours in the y axis? That the corresponding task takes N hours to humans? To AI? Why don't just explain things clearly?
>>
>>108348723
>Can't measure 2025 in 15 hours
Guys is it over for me?
>>
>>108348635
where is gpt codex?
>>
>>108348784
Still being evaluated... hopefully it sucks, please pray with me
>>
>>108348746
It's gobbledygook
>>
>>108348793
if it says that it's better than Opus 4.6 benchmark is a meme.
>>
>>108348809
AI benchmarks have always been memes. They're made up marketing tools whose only purpose is to make the new model look exponentially better than the old one regardless of the actual change.
>>
>>108348723
Since you think that correctly interpreting a diagram without ANY FUCKING CONTEXT has anything to do with the ability to operate a computer, I'll report the ACTUAL caption from the ACTUAL source:

> The length of tasks (measured by how long they take human professionals) that generalist frontier model agents can complete autonomously with 50% reliability has been doubling approximately every 7 months for the last 6 years[...]

This part is ESSENTIAL to understand this picture.

> t. math graduate
>>
>>108348831
If its not clear to you what the graph means already then you dont need to worry about AI because there was never anything to replace

> t. math phd (formal logic)
>>
>>108348809
Prediction markets seem to agree with you. President of openai tweeted something about how "we dont need benchmarks where we're going" which makes me excited that you might be right.

I need opus 4.7 to barely improve because that curly haired freak dario is most giddy about the idea of destroying our livelyhoods, and so far, also most successful
>>
>>108348852
Yet the ACTUAL AUTHOR thought that it was necessary to put a FUCKING caption, like any other scientific paper ever published.

> formal logic
a meme subject, second only to set theory. Formal logic is a way to say "I'm too dumb to grasp algebra, geometry or analysis".
>>
>>108348635
Just 2 more time units.
>>
>>108348635
I had a hypothesis that although LLMs can't do everything humans can do, it will be able to do a certain few crazy things that NO human can do, and it's looking like many more things like that are coming than I thought.
unironically over.
>>
>>108348940
I was also a sceptic... i really need it to fail soon
>>
File: drdaffy.png (564 KB, 719x719)
564 KB
564 KB PNG
>>108348635
>increases estimate on the current task from 6 to 12 anyway
not my problem
>>
File: 10987654256347685.jpg (31 KB, 460x501)
31 KB
31 KB JPG
>>108348635
what is the methodology?
- do they just attach a txt file with code in it and tell model to identify used language, check for bugs, and fix if any found, or
- they do fuckton of shenanigans with prompts spoonfeeding the model exactly what to do and where?
because it does matter. if it can fix "complex bug in ml research codebase" - why it cannot fix 20 y/o bugs in looneks codebase? certainly, "ml research codebase" sounds like a much more serious shit than some of those old-ass bugs known to everyone yet never fixed.
>>
>>108348635
Are you using the vibecoded browser?
Are you using the vibecoded operating system?
Are you using the vibecoded programming language? (No one has done this yet, but someone probably will)

That's what I thought.
>>
Hey, OP here, I was wrong: https://www.transformernews.ai/p/against-the-metr-graph-coding-capabilities-software-jobs-task-ai

We're okay for a few years yet. Remember to save and invest

>>108349060
Theres loads of vibe coded programming languages at this point
>>
>>108349060
>Are you using the vibecoded browser?
Google is pushing ai hard internally ao soon everyone will.
>Are you using the vibecoded operating system?
Windows is already vibe coded and torvalds is pro ai.
>Are you using the vibecoded programming language?
probably never will.
>No one has done this yet, but someone probably will
What do you mean? We are getting new vibecoded language every week. Not even the creators are using it though.
>>
File: HDIGa4HWsAA6mq_.jpg (170 KB, 1360x1490)
170 KB
170 KB JPG
>>108348635
Total Opus victory.
>>
File: amazon.png (365 KB, 1448x766)
365 KB
365 KB PNG
>>108349126
Amazon has been pushing AI internally and look what happened

https://arstechnica.com/ai/2026/03/after-outages-amazon-to-make-senior-engineers-sign-off-on-ai-assisted-changes/
>>
>>108349605
Amaslop has the shittiest engineers
>>
>>108348635
can they just improve memory parsing so I can have slow burn romance RPs with the bot for longer before the model starts to rot?
>>
>>108348831
you don't need to be a math graduate to understand why the caption is important. just write a single fucking paper in a pregrad class

>t. micriobio graduate
>>
>>108348635
>we have a tool that can automate tremendous amounts of work
>this means that "it's over"
k



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.