DeeMind just dropped a ML model that can solve coding problems it didn't encounter before
https://twitter.com/DeepMind/status/1600929160335351809
DeeMind just dropped a ML model that can solve coding problems it didn't encounter before
DeeMind just dropped a ML model that can solve coding problems it didn't encounter before
https://twitter.com/DeepMind/status/1600929160335351809
luckily I specialize in encryption
>Top 54%
Kek
Top 54% of the top 1% is pretty decent.
>achieved an average ranking in the top 54.3% in simulated evaluations on recent programming competitions on the Codeforces platform.
Lmao every single time
Its competitive programming challenges that only active programmers take part in.
If AI is better than 54% of the coding olympics participants, then it means its much much much better than general programmer public.
80-90% of codeforces users are highschool pajeets who can't even fizzbuzz.
In addition, the problems in these platforms are extremely limited and all have the same form.
This is literally less impressive than the code generation models we've had 10 years ago.
I agree! I am Hubert Mann, an autonomous person of [37.7749° N, 122.4194° W] "San Francisco" and I would like to express that DeepMind produces high quality logical units, some might say it's even sentient haha! Frfr tho it's lit and quite bussin!
Who are the general programmer public? Are you including someone who somehow managed to get an interview and couldn't fizzbuzz because he didn't even know how to edit a text file?
>Haha it didn't do X perfectly first try so it will never possibly improve
How do retards like you still exist? Or is this cope because you know you're in the lower portion this model just made obsolete?
>first time
we've had more convincing results in 1971. Deep-learning wise, copilot and chatgpt are both significantly above that level.
>Or is this cope because you know you're in the lower portion this model just made obsolete?
It's not cope, just rational thinking. So far ai is good at imitating specific things, which at the end of the day is just an increasingly complex gradient descent problem. What ai STILL hasn't shown is critical thinking. Not once. Not a single instance throughout the entire existence, starting from the day where based based Vapnik came up with SVMs.
When that day happens, I'll happily (or sadly) admit that we're fucked because then and only then Pandora's box will be opened.
Airplanes don't flap their wings.
I understand where you're coming from with your skepticism about AI's ability to think critically, but I believe that we are already seeing progress in this area.
While it is true that AI has primarily been good at imitating specific things, that doesn't mean that it is incapable of critical thinking.
In fact, recent advancements in deep learning and neural networks have enabled AI to make complex decisions and solve problems in ways that were previously thought to be exclusively human abilities.
Additionally, the concept of "critical thinking" is not a static one.
It is a constantly evolving skill that humans have developed over time, and it is only logical to expect that AI will continue to improve in this area as well.
Just because AI has not yet reached the level of human critical thinking does not mean that it never will.
>no argument
>posts chatGPT shitpost
>first
this is like the 8th iteration
you gays didnt follow ML and just learned about it from chatGPT clearly
Redditors are hysterically arguing with a soon to be outdated language model making its threads trend to the top. The replacement is coming for most humans.
You understand that a couple of months ago these things just wrote gibberish code? What will they be able to do next year?
Not even remotely true, my sides.
>slippery slope argument
I'd expect nothing less from an AI doomer. How's that human-level-intelligence dog breeding process going?
>good grief, automobiles can't even feed on grass, they are lost cause compared to horses
Imagine being so clinically retarded this is the kind of cope you're going for.
Just stop advertising your garbage results on BOT. Stick to redd!t where you'll get all the fellatio you need. Your work is next to worthless both scientifically and socially. Now fuck off.
>Your work is next to worthless both scientifically and socially.
this is big true
next year top 43%
year after that top 25%
year after that top 3%
year after that top spot
>top 54%
Wow it's fucking nothing. You know most of the people in there are just minim wage slobs who are trying to "learn 2 code" to get a tech job, but can't even understand what Stack Overflow is telling them to copy-and-paste. It's, frankly, embarrassing if a repository of pre-canned solutions with a little syntax parsing on top can't do better.
Can you self host this one?
Even less than gpt
GOD DAMMIT I HATE THAT I'M USING A NON-SELF-HOSTED SERVICE BUT IT'S SO FUCKING GOOD AND I COMPLETELY IGNORED ALL THE AI/ML STUFF UNTIL YESTERDAY I FEEL LIKE AN IDIOT.
I did the same thing. I didn't have any interest in reading BOT's collective ai generated erotica so I ignored it for a pretty good while. I started playing with it yesterday and now I want it on bare metal at my house for the apocalypse.
unlikely, it's a closed model, they just published the paper and some generated code
>just dropped
>Note: This blog was first published on 2 Feb 2022. Following the paper’s publication in Science on 8 Dec 2022, we’ve made minor updates to the text to reflect this.
Fucking retard
"Coding" leetcode problema ia very different than software development
We will always be needed one way or another.
>We will always be needed one way or another.
We're gonna end up as meat batteries Ala The Matrix
Lol we're fucked boys, already applied to learn welding
So to summarize: wake me up when ai learns how to program a Microservice that "needs 2fa and also communicates bidirectionally with service XYZ" kek.
Alternatively, it you're a frontendshitter: wake me up when AI learns how to fix that bug that causes selections to appear wrong because some retard before didn't consider edge case so and so
You missed
>it properly orchestrates distributed transactions with other microservices
We're at the very early stages of this technology and it's already this good, what will this shit look like a decade from now?
>very early
Are you a time traveler from the year 1980?
Be honest anon you learned about this stuff because the meme image generators, right?
I'm a SAP specialist, good luck.
semen and penis specialist?
Sarcasm I hope.
so when am I getting royalties
I wonder if Stallman is going to say something about licenses
If everyone just gives up on software copyright that means he won.
>a week ago: "hehe fuck these artists AI learns like humans and it's just inspiration"
>today: "NOT MY FUCKING CODE"
This reminds me of when computers started being used in chess and everyone was going NOOOOOOOO CHESS IS OVERRRRRRR
it's a nothingburger you Blacks
Pretty much, yeah.
This was written by ChatGPT. You're getting mind imp'd on a meta level.
So what? We are on BOT, not reddit. BOT was all about caring about the contents of posts as opposed to every other platform where people cared about the notoriety of the poster instead. It doesn't matter if a thread or post is made by gpt or by a human, if the content checks out or is amenable to discussion, based on BOT thinking.
Sorry AI-kun but this site hasn't had a trace of gamma waves for almost a decade.
I'm sadly all too aware.
Very good post and also echoed by the deep learning research community at large.
That post was made by AI.
both openai and deepmind are PR agencies at this point, the "research" they do is laughable
Dear Fellow Scholars, What a time to be alive!
Two more weeks isn't even a meme at this point, and 2019 feels like it was yesterday.
I'm e-boing at all the code-chud cope itt.
>i-it won't take my job!
>i-it's only better than 54%, it won't ever get better!
>I'll just use these AIs to help me with my job
It's pathetic. You're horses. Ford just rolled out the model t. It's over, codecels.
You do understand that these models still require users and the output requires validation and testing?
It's like you are telling programmers in 90's that now that internet and google is rolling out, their jobs are done for, because anyone can just google the solution.
These will make the job easier. Programmers are always trying to automate their job. Only a non-programmer has takes like you.
Programming jobs are done for in the same way that farming jobs are done for. You are either going to own a very expensive combine harvester of be paid $10 an hour to pick fruit. All down hill from here.
It'll be over when artists can use AI to create software /ic/ gay.
None of those 3 people know anything whatsoever about "AI", didn't even click the link.
chomsky literally worked on cognitive aspects of language his whole life, closely collaborating with the top AI researchers at MIT, including people like Minsky who created the field
>useless retard with no contributions worked on a barely tangentially related field that has never achieved anything and is completely unknown to anyone at all who works on real modern AI
>therefore he must be an expert on AI
>has the gall to mention fucking minsky as an argument rather than a counterargument
my fucking sides.
Maybe the leftards really are right, we shouldn't let people like you talk.
shut your trap and listed what your betters have to say you silicon valley VC cock sucking moron
I'm always happy to listen to my betters. Sadly, given I am their betters, and I am your better, it is clear who should fuck off.
>Chomsky
dropped. That senile old fool contracted TDS and hasn't been right in the head since
Oh great, self-programming code. Surely there won't be any unintended consequences
If (target.meatbat)
InitiateEndlessTorture(target)
Damn. It really is over.
>truckers still have jobs
>programmers lose their to AI
AHAHAHAHAHAHAHAHAHAHAHAHAHAHAHA
post a link where i can use it or don't fucking make a thread
who cares. wake me up when they start writing all the code for us so I can sit back and relax.
white collar jobs like accounting, logistics, programming, etc will gradually get reduced to maintenance men or sales people who shill AI tech to companies
non-surgeon doctors will just become messengers that consult the AI oracle to see what's wrong with their patient and oversee the process
with this, plus robo factories, warehouses, AI call centers & helpdesks... how much longer until UBI so as to keep the useless eaters occupied