please do not give them any new ideas

Free will under determinism is the most basic midwit filter.
>These aren't compatibility so determinism is false
Midwit
>These aren't compatible so I don't have free will!
Retard with delusion of intelligence
>These are compatible
High IQ

based tech mommy

>>102764297
>grifter telling me other grifters are finished
no thanks

>>102764297
remember that time she made vid about global warming being sponsored bullshit and not really rigorously researched which made so many (((people))) outrage she had to backpedal it in later vids

>>102764625
yeah that's what scientists do. they re-adjust their worldviews if they find out they're wrong which is what she does. compared to your type that just have a larger and larger tantrum when they get presented with evidence that they are wrong

That's surprising, I'd think someone who is all for no free will and representation of the brain algorithmically would be the first to say AI will be sentient eventually.

>>102764546
Compatibilism is literally "it was real in my mind".

I can save her

>>102764297
hey Sabine, when are we going to have that bait and switch moment? pretty much everyone else already sold out. do you not have rent to pay or something?

>>102764658
its just a """coincidence""" you literally cannot discuss anything about it which goes against general consensus or point out problems with sponsored research otherwise you get deplatformed or fired. Similar with covid and doctors who openly disagreed with actions of politicians who were clearly taking bribes from pfizer, i guess thats how science works nowadays.

>>102764877
>nowadays
It was always the case. Internet just made things like this easier to notice, hence the bias.

Sabine is the Technology Connections guy if he were based and not cringe.

>>102765233
Alec Connections > Muh Determinism lady

>>102765233
Sabines cringe as fuck lmao.

>>102764761
She's a normie, she has a family and kids and had a stable career that she gave up because she wanted to do something more fulfilling.

Unlike you faggot

>>102765233
nothing worse than scientists dropping hot opinions on things they know nothing about and acting like they should be taken seriously because they have a doctorate.

>>102764297
Yeah that's right it's not the future of coding, it's the present

Sabine really is the ultimate contrarian.
>academia: bad
>climate change: fake
>ai: bubble
There’s certainly room for criticism here but it’s so obvious she’s using her PhD to come across as an authority figure and choosing edgy opinions for views.

>>102764297
It has only been 2 years since the AI boom, AI algorithms are clearly still in its infancy considering our brains can accomplish so much more than todays AI with less processing power than a gigacomputer with thousands of GPUs
There will be another breakthrough and people will finally learn that computers can surpass brains in nearly every single way. I only hope that the people who are so adamant about the mythical status of the brain live long enough to eat their words

>>102766568
She literally said that eating the bugs is the future and climate change isn't being feared enough

>>102766692
how is the brain a computer? what does it compute?

>>102764877
>you literally cannot discuss anything about it which goes against general consensus or point out problems with sponsored research otherwise you get deplatformed or fired
because you're not capable of civil discussion. as i said before when you're pointed out that you're wrong, you have a tantrum and explode into a fit of racial epithets.

File: 1728583128063009.png (380 KB, 547x867)

380 KB PNG

>>102766568
>choosing edgy opinions for views
mfw

>>102767180
Are those nipples also for views?

>>102764546
>One event causes the other
>Your thoughts and behavior are a result of events out of your control
>Somehow you still have control over your actions

Excuse my ESL ass, but how the fuck does that work?

>>102765233
Are you stupid?
Sabine talks about theoretical physics. The other guy is just obsessed with, like, antique toasters or other mundane shit

>>102764658
social pressure and shaming isn't scientific correction you tranny retard

>>102766962
Define how

>>102767436
External factors increase or decrease the chance of certain outcomes, there is never a guarantee. Events out of your control will determine the pool of most likely outcomes, but you are still making decisions within your new more narrow selection. And there is a chance you will choose to do something statistically unlikely.

>>102764297
is that a troon

>>102767180
>sub mid but biological female
>3/10 elsewhere but 7/10 among contrarian pseuds
>dresses like a whore to attract views
Many such cases

>>102767942
no, just a german.

File: bingbing.jpg (51 KB, 1113x709)

51 KB JPG

Once again a boomer coper uses vague generalizations and unfalsifiable claims (AI thinking != human thinking thus AI can never be humans!) to justify a feeling they already have.

The study she cites makes no distinction on the version of the model the programmers were using. It only says "Github Copilot", which could be based on GPT-3.5 (garbage), but whether it is makes no difference, people using Cursor + Claude 3.5 profess universally that it's miles ahead Microsoft's half-baked Copilot cash grab.

This is the essence of the cope. People assume that the current capabilities of models is the best they'll ever be, and even if a model exists today that does something they say LLM's could never do, they erroneously cite some study that uses some old, lower quality model to affirm their preconceptions.

LLM's are based on a technology that has existed for only 7 years. Most of what we are seeing now are proofs of concept that show that the techniques and investment are worthwhile. OpenAI, Anthropic, etc. know that the prize isn't GPT-4 + some shitty tools, it's AGI. All the current models are just available publicly to they can drum up investment hype so they can get funding to pay for their training of the really big models that can actually do useful stuff.

If you see an AI do something people said would be impossible for decades mere months ago, then focus on some mistake it makes and think "haha dumb AI doesn't know (thing obvious to humans) it will never be as smart as humans", you are doing something called "grasping at straws".

>>102764297
She's not a coder
She's not a AI researcher
She doesn't work in the computer industry

She's a youtuber e-celeb without expertise, yapping

>>102764297
>judean grifting bitch

File: thesundestroyingeverythin(...).jpg (79 KB, 1113x709)

79 KB JPG

>>102768134
There was a slight inaccuracy in the image I thought needed correcting.

>>102768134
>LLM's are based on a technology that has existed for only 7 years. Most of what we are seeing now are proofs of concept that show that the techniques and investment are worthwhile. OpenAI, Anthropic, etc. know that the prize isn't GPT-4 + some shitty tools, it's AGI. All the current models are just available publicly to they can drum up investment hype so they can get funding to pay for their training of the really big models that can actually do useful stuff.
So it is literally burning money because LLMs aren't a path to AGI and everyone fucking knows it.

Jetbrains' line by line code completion is pretty nifty, but that's a local model and works off of your code. It's like a supercharged intellisense. Everything else seems gimmicky for development.

>>102764546
Determinism isn't real, retard. It's disproven by quantum physics decades ago. True randomness exists and is inherent to the fabric of reality.
>>these aren't compatible so determinism is false
Would be a retarded argument, but that's not why determinism is false

>>102768361
>So it is literally burning money because LLMs aren't a path to AGI and everyone fucking knows it.

lmao they are a stepping stone to AGI and you're coping if you think you're right and all the 150IQ researchers at top labs are wrong. All tech companies operate on a loss early on because of the competitive environment and need for upfront capital.

Why would billions be invested in a prospect that "everyone fucking knows" won't work. You really overestimate yourself if you think you're smarter than Meta, Google, OpenAI, Anthropic combined. But go ahead, enjoy not making high 6 figures working at these places while you insist it's all a big dead end

>>102768134
I'm interested in what the area outside "what a human can do with a computer" represents. What are some examples of that?

>>102768483
>Why would billions be invested in a prospect that "everyone fucking knows" won't work.
Because they're not investing in AGI, they investing in chatbots which they think will improve productivity or reduce staff coats.

>>102768570
I agree your sentiment, but I'll note that developing AGI is the stated goal of these companies.

>>102768584
That's fine, but AGI is not going to be an LLM.

>>102768134
the current approach was brute force. But it's getting hard adding fuel to the fire.

>>102768483
>Why would billions be invested in a prospect that "everyone fucking knows" won't work.
The idea of not paying people for work and making more money makes the jews wet their pants so they invest. It's literally hopium.
Researchers will do research gladly because they are interested in the subject and/or money. They have money so they do research.

>e-celeb thread

>>102768410
Please don't embarrass yourself, the mathematics of quantum mechanics show the probabilistic nature of reality and the evolution of any system is 100% deterministic and continuous according to the Schrödinger equation . It is only when you invent wave function collapse outside of mathematics that you can maybe talk about randomness in the universe, there is certainly nothing within the mathematics of quantum physics that disproves determinism.

>>102768410
>True randomness exists and is inherent to the fabric of reality.
randomness is a man-made concept. just because our monkey brains don't understand it yet doesn't mean it doesn't have some sort of order behind it.

>>102764297
>I didn't believe AI is the future of coding
Stopped reading right there, how can you pretend to understand technology and not believe AI will replace programmers ?

What an absolute tool.

File: cfdd56b26ade080449070dd03(...).jpg (75 KB, 1104x580)

75 KB JPG

>>102768862
>the probabilistic nature of reality is 100% deterministic

>>102768410
>True randomness exists and is inherent to the fabric of reality.
So? How can free will come from stochastic processes either?
>It's disproven by quantum physics decades ago.
It hasn't been proven, the de broglie-bohm interpretation of quantum mechanics is as mathematically valid as the copenhagen interpretation. It's not possible to devise a test to determine which one is true - the universe might be deterministic with nonlocal processes or local with stochastic processes.

>>102767874
>And there is a chance you will choose to do something statistically unlikely.
That's not free will, that's floundering. Is fish exercising free will when they flop around out of water? Because that's what muh making decisions is.

>>102769320
Uh, no, that's literally not true. You don't need more than a high-school-level education in physics to know that in quantum mechanics some things are literally random. There are no hidden variables behind it like you said; it's genuinely indeterminate until measured.

>>102769464
Superdeterminism (nonlocal determinism) implies time travel (the future causes the past).
That's why it's usually rejected by mainstream Physics.

>>102768862
>>102769881
The Nash Equilibrium for a game of Rock-Paper-Scissors is to play a Superposition of all three states with equal probability.
Without Wave-Function Collapse, you can never finish a game of Rock-Paper-Scissors.

>>102765233
Technology Connections guy is based.

>>102767180
shame she never had her eggs fertilized
those nipples could have fed many healthy babies

>102768410
>102769963
>102771608
all of these, and most of modern quantum physics, assume chsh inequality, which is not nearly as firmly established as most people believe
in fact, the entire ch74 experiment, which is the only experiment design that "proves" bell inequality, can be dismissed if the particles are losing decoherance before they reach the detector
and since the "slit" is a quantum object made of quantum particles, i find it impossible to believe that they wouldnt lose decoherence
physicists cant admit this because it would admit to 50 years of shame and fraud

>>102768550
It's mostly speed, as in LLM's can write far faster than humans. They can perform semantic analysis of a large piece of text and compose semi-decent poetry with difficult constraints in minutes that would take hours for average humans.

The current OpenAI voice mode knows more languages than any human.

>>102764297
>youtuber: it's over
>Google: *makes an LLM use LEAN to write proofs and win math competitions*
>it's just autocomplete! useless useless autocomplete!!!
>OpenAI: *makes a model correct itself and win codeforces, math competitions*
>IT WILL NEVER CODE
>Anthropic: *releases a free mid-sized model so good at programming it singlehandedly makes VSC-LLM integration devs millionaires*
and the best model for THIS YEAR isn't even out yet.

Anons, this is a sincere post for once: stop being in denial. We are FUCKED.
>AI won't replace people, people using AI will
yes and those people are the fucking Indians doing it for ruppees. I repeat, we are fucked. Unlike the boomers in trucking and other trades, most of us don't have unions.
Please get comfortable with the idea that even the laziest shut-ins in this board will have to go out to the streets and protest for UBI, against offshoring, or some sort of solution. They're not just going to hand it to us. The job market is already fucked without AI, just imagine how bad it'll be in 1-2 years.
>>102768134
I wouldn't mind so much if it was just old youtubers with no skin in the game commentating about shit they know nothing about. But no. All my dev friends are in denial. No one's preparing, no one does shit about it. Look at the longshoremen and how they put their foot down for job security. That could've been us. But most tech autists are too self-aggrandizing and smug to admit they're replaceable cattle, so now the big corpos are ready to steamroll all of us with zero resistance.

File: 1443754900613.gif (999 KB, 500x281)

999 KB GIF

>>102764557
this

>>102771697
this I was worried he might be a faggot, but turns out he's married, so hes based again

>>102769881
Then why do you think determinism has anything to do with free will? If making random decisions doesn't make you free, then whether the world is deterministic or not doesn't matter when it comes to free will.

>>102764546 is right in the sense that free will and determinism are completely transversal questions.

>>102764658
sure so science is not even about truth nor knowledge, it's just about the convention of a tiny group of people who use cryptic language to ''talk about models''

>>102764297
baseded
I'm enjoying watching the "AI" meme slowly crumble

File: 1726845847210626.jpg (52 KB, 1113x709)

52 KB JPG

>>102768134
fixeded for u

>>102769435
He's right though, the schrodinger equation is deterministic. The "interpretation" may or may not be, it's an open question.

>>102764297
This hag is straight up spreading missinformation, she literaly doesn't know.

>>102766938
technically, athropods are insects isn't it?
I like prawns

>>102772817
She's completely right though. I'm sorry that your dreams of quick success without mastering any skills are crumbling, Pajeet-sama.

>>102764297
>>102772989
I want my "I was right" award, but I never receive those even though I'm not a snarky douche-bag about it.

>>102773246
just use AI to generate one for you luddite.

File: 1698104027622611.jpg (704 KB, 2160x2923)

704 KB JPG

I don't know
o1 is a massive improvement over 4o
Honestly kind of scary
I'm just hoping there'll be UBI so I don't succumb to the elements in my tent

>>102771716
she is married with two children.

>>102773282
>o1 is a massive improvement over 4o
Lol. Most of the "improvement" in ChatGPT over the past years has been in supporting code, like tokenizers, rather than model improvements.
They pour exponentially more items into the language models and it barely changes.

>>102768134
>Most of what we are seeing now are proofs
That it's shit. Only people who are winning here are those who like chat-bots.
Also neither LLMs or any predictive models are an actual novelty. The theory been around forever and "muh seven years" is just astroturfing the recent commercial faggotry.

>>102767874
You will never do anything unlikely. That's just an epistemic possibility, metaphysically you can only do 1 thing. The statistics you talk about just model that epistemic uncertainty.

>>102764546
>two binary states
>not considering the 4th option
Clearly free will is an illusion, and determinism is a lie sold to make people feel safe.
Nothing is true, and nothing is permitted.

>>102773303
So she's a pedophile bigamist, what a sick fuck

>>102772989
>I just want to believe what I want even if reality is completely different
All that retard said is misleading, even local models are already decent. Most of it is addressed here >>102772094 >>102768134
That hag is absolutely retarded and a grifter, most normal people actually using AI knows x100 more than that retarded hag do.
Is only going to hurt you in the long run, retardo san.

>>102764297
that's a man

>>102773351
>Lol. Most of the "improvement" in ChatGPT over the past years has been in supporting code, like tokenizers, rather than model improvements.
you're utterly clueless. gpt4o sucked ass on release and now it's on par with Claude Sonnet 3.5 (which is a huge deal cause Sonnet 3.5 is LEAGUES ahead of even GPT-4-013). o1-mini and preview are shattering benchmarks. Opus 3.5 isn't even out yet.
>They pour exponentially more items into the language models and it barely changes.
wrong. People really want this to be true but the scaling laws still hold in 2024.

>>102773414
>do thing
>you were just likely to do that thing bro
literal gambling fallacy

not a single good coder I know uses LLMs. all the ones that do are pseuds who started coding way into their twenties to get muh job.

>>102766568
i stopped watching her when she did a whole video about how 2 + 2 can equal 5 or some shit. it wasn't even a clever thought experiment, just a bunch of mental gymnastics.
she can be decent at times but most of her vids are pointless slop.

Why does her thumbnail look like a nintendo cartridge / playstation cd case?

Looks like I am playing cory in the house: crazy nanny for the sega genesis

her tranny video was too cringe for me

>>102773958
>Looks like I am playing cory in the house: crazy nanny for the sega genesis
fucking lol. can't unsee it

>>102771845
This sounds schizophrenic. Where can I read this critique of bell?

>>102764297
coding, meaning maths and logic is literally one of the only things that AI will most definitely be the only future of

>I said make me a new video game in C++ with chatgpt 3.0 and it didn't work!!! AI IS OVER!!!

>>102774027
yes, coding AI is definitely the future. along with time travel and cold fusion, kek.

>>102774054
ofc it is.
coding is using the same 20 principles and same 40 modules over and over again to achieve the same goals over and over again.
LITERALLY one of the only things AI will most definitely replace

>>102767436
It's basically a technofag version of Calvinism.

>>102764297
Can't stand this insufferable, hypocritical, intellectually dishonest cunt.

File: lookimcooding.jpg (63 KB, 737x499)

63 KB JPG

>>102764297
Cope, jews will replace you

>>102774143
If this was true then outsourcing to third worlders being paid 10 cents an hour wouldn't literally always be a disaster, and yet, it is

File: altman.jpg (97 KB, 799x930)

97 KB JPG

>>102774143
we'll live to see anon, but given how much there is for big tech to gain by dropping coders, the progress is shit slow, and it looks like we have already hit the wall. some sci-fi AI may replace coders in the far future, but there is nothing out there now that could do that.

I've never seen more than a few seconds of her video because she comes across as a cunty, german, headmistress

>>102773946
You don't know a single good coder

>>102764546
(I don't care if randomness exists or not, it has nothing to do with free will.
Particles move according to laws of physics so there's no free will, or in other words, you couldn't "have chosen otherwise". The situation you were in led to you making the choice you made, any hypothetical other choices are just delusions and "what ifs" in your mind, and not real "timelines" that could have actually happened.)

An idea I never see people talking about is that, you can still attribute choices to people in a way that appeals to common sense, eg. you went through a decision-making process where you pondered between multiple options in your mind and ended up choosing one (even though the ingredients in the process were outside of anyone's control), so the action now "belongs" to you and so by human morals, this makes you responsible and accountable for it.

It gets more blurry when it comes to overt manipulation, where if the causal links are strongly attributed to a more isolated source instead of a varied environment, we tend to think the action is then less attributed to the actor. But in the end this is just a man-made concept to appeal to our emotional sense of agency and morality. It's a rough estimation of causality to help diagnosing and fixing most obvious problems in society. But in the complex and chaotic environments that reality is, most thoughts and actions are practically impossible to predict and intuitively explain. With the use of science and big data, and now unintuitive statistical algorithms, we have become quite a bit better at it.

The underlying purpose of Al is to allow wealth to access skill while removing from the skilled the ability to access wealth.

File: 1727948559110433.jpg (35 KB, 680x667)

35 KB JPG

>>102764297
she's just mad they gave the Nobel Price in Physics to Geoffrey Hinton specifically for starting the modern AI revolution
I like Sabine but every time she talks about what's going on in the generative AI space I roll my eyes, she's not an authority figure in the field and she quite frankly doesn't know how it works on a technical level

>>102774983
shes right though

>>102775039
it entirely depends on what the end goal is
is it to have a sentient, conscious AI? well we can't even verify a human's consciousness so this is a moot point, but most companies aren't interested in this anyways

is it to provide a customer-facing service? this clearly has users but financially they don't know how to turn a profit yet, and my guess is its a long road of monopolization by Microsoft of OpenAI and others before enshittification kicks in

is it replacing human labor? this is definitely winning out, there's no stopping it. its happening b2b (business to business) but still isn't super sustainable for openAI. However, companies are realizing you can just use local models on-premises (and anyone that cares about security should do this anyways) and massively cut your own costs

being an AI provider is not yet profitable (and may never be) and API rates are artificially low now for sake of killing the competition, but being an AI user on-premises is feasible today and saves millions in human salary if not more depending on your company's scaling

File: Untitled.png (792 KB, 1024x1024)

792 KB PNG

>>102772783

>>102775124
no need to go that far
if a couple pixels are off in an image, nobody will notice
if your program changes an assumption halfway through it doesnt work
its not only a matter of precision, its also a matter of context size
for instance, if half your program works with the assumption that a pixel is rgb, and the other that a pixel is argb, your program wont work either
its all a matter of maths:
as the context gets larger, you need exponentially more computations to create a model.
if you want to increase your precision, you need exponentially more data
thats the true limit
it was the case since the beginning because that limit is inherent to the technology

>>102775180
at its current stage, it operates as a junior "employee" that needs a human senior to babysit it
but its much faster than an actual human junior
that eliminates like 70% of human labor off the bat - juniors and junior-level employees

the reason for example new grads are struggling to get hired isn't about the interest rates alone - those are already being cut, but hiring isn't going up, because there is no use case for juniors any more

>>102764330
>ai coding
>new idea
???

>>102775217
nonono
stop with the antropomorphization of mlai
realize that at the root of it are the same maths you would use when you do a market analysis in finance
things are not a spectrum
theres exact specs to things which make mlai suitable for certain jobs and totally worthless in others
but in all that
you dont get 100% precision so ml wont replace jobs, properly speaking, it will accelerate existing ones.
which will reduce the workforce. so depending on your personal semantics, it could be kinda misconstrued as replacement.
just as much as the calculator replaced engineers

>>102768483
Its not engineers that are the most optimistic about ai development thought. Its mostly that investors that believe llms to be so revolutionary, so they are willing to give capital to credible companies that develop them. Google, Meta, OpenAI etc are just there happy to collect their funds and grow

>>102775281
it is currently replacing new grads and juniors
we're already seeing this
yes its like a calculator for seniors, but only for them because new employees don't have the experience to make corrections

so the replacement is already here
instead of senior A assigning tasks to juniors B, C, and D and doing review and fixing later
we have senior A assigning tasks to generative AI models X, Y, Z, and doing review and fixing later
that's 3 jobs gone with a couple of GPUs

>>102775399
yeah no
thats not why you get layoffs
you get layoffs because most people are completely worthless

if you want to see advances in the industry using ai you have to look entirely elsewhere
https://www.youtube.com/watch?v=Cj-bm7zt0bI

and it isnt even the same type of ai.
its symbolic ai
and obviously its not marketed or hyped and such
bc why would you give an edge to the competition?
ai assisted design is nothing new tho
you could call using a spreadsheet- ai assisted design

>>102775456
>you get layoffs because most people are completely worthless
and easily replaced with generative models
I'm not sure why you're splitting hairs. I'm not speculating here, I'm *telling* you this is already happening. Every major company is doing it including the one I work at. $20k in on-premises compute goes a long way and sales millions in salary.

>>102775162
kek

>>102775495
*saves millions in salary

>>102775399
layoffs have nothing to do with AI

>>102775529
>>102775529
layoffs started because high interest rates made the old model intractable, yes
but interest rates are going down, and full-time hiring isn't
its precisely because there's an alternative to human labor now that this is happening

>>102775590
*full-time hiring isn't going up

File: Screenshot from 2024-08-1(...).png (90 KB, 709x651)

90 KB PNG

>>102775495
>i'm *telling* you this is already happening.
if you can actually replace people with this garbage
you must be a webdev or something
and your company must be veeery inefficient

the code is fucking worthless otherwise
its literally unusable
here, i got an example
i dont have it in full but count_vector and its associated operations are not used later on

its complete fucking trash if its not regurgitating something you could find on stack overflow, and close to verbatim anyways

>>102766692
>AI algorithms are clearly still in its infancy
bro the AI boom is literally putting severe strain on the American power grid, this shit is dead end technoloy, it won't get any better than what it is now.

>>102775602
not webdev, in an engineering field that is very heavy on physics actually
again, you have to manage the output, you're completely misunderstanding what I'm saying (I think on purpose)

you as a senior can easily fix the problem, but it would have taken a human junior possibly a day or more to write this (a week if its new stuff to them entirely), while this basic template was generated in seconds
you make your small patches and you're good to go

>>102775632
>very heavy on physics actually
gotcha
>using ML to do eng calculations
i call larp

>>102775632
also if it would take a junior more than a day to wrap their head around vectorization they have nothing to do in my company

>>102775632
Anon all this "AI" chatbot garbage does is regurgitate blog posts and stack overflow answers written by Indians who don't know what they're doing. That's all it can do. It has no understanding of anything, it's just generating plausible-looking text.

>>102775642
who said I was doing engineering calculations with ML? again, you are purposely fucking around

as a senior dev, I have planned subsystems for an architecture
I define what each subsystem has to do rigidly, and come up with interfaces and behaviors
I would then assign this spec to implement to a junior, but now I can just give it to models
the model automatically checks in to a branch and I do code review

I'm not asking a model to reinvent matrix inversion or the FFT, I'm asking it to write code

>>102764297
I imagine they're go back to shilling shitcoins.

>>102775661
Did you remember to also search for the code it "generated" verbatim to see if the github repo it stole it from was GPL licensed?
I write embedded software and we have a company-wide hard "no AI generated code" policy, partly because it is extremely good at generating code that looks correct but is subtly wrong and would kill somebody, but also because it's impossible to tell if the output is infringing on copyright

>>102773951
>most of her vids are pointless slop
>most of her vids are doom porn
ftfy

>>102775661
hopefully that works for you
because it wouldn't for me
in 99% of the cases its just easier to rewrite the whole code than to first learn it, and then debug the fucking mess
and yes, i dont work with juniors. were actually a small team of pretty competent people when im not soloing my code

>>102775694
Doesn't matter, we're a subsidiary of Microsoft and they'll bury anyone that comes after us in counter suits

AI literally is the future of human evolution. It is the next step in evolution and once that happens, we ordinary biological humans will become obsolete. That is entirely correct and a good thing in my opinion, we are useless retards.

>>102775655
I mean, if you can learn programming by googling everything and reading stack overflow answers (or similar resources), then the AI has all the information it needs to become a capable programmer. Most people learn by copying others, and most ideas you have thought of already exist on the internet, truly original ideas are incredibly rare.
I mean, I wouldn't trust AI to program a video game for me, but basic codemonkey jobs are going to be obsolete.

>>102775162
Underrated.

>>102775711
aaah now i think i get it
>classification problem larping as a generative one
the majority of your code has been done to death

bc when you ask mlai to generate code, proper
its just shit the bed with bloody diarhea

>>102775727
the majority of the code anyone writes has been done to death, in actual products the plumbing is 90% of the codebase and is uninteresting

like I said, we're not reinventing math with generative AI, we've already done that work

but even if we were, a junior would be useless in that field anyways

>>102775590
it takes at least 6 months for lower rates to have any effect on the market. it's too early to tell anything. also, we might be heading for a full-on recession, so don't expect job market to get better any time soon. still nothing to do with AI.

>>102775750
when we hit a recession and companies still need to deliver YOY returns to their investors they'll reach for AI and management everywhere will realize most new grads are a useless timesink

>>102764297
>Female Bill Nye
No thanks

>>102775717
I think the sign of a good programmer is if every time you have to resort to searching stackoverflow, the only result is one guy asking the same question 4 years ago with no answers
You can't learn how to solve problems by copy/pasting code fragments and neither can "AI"

>>102775742
>a junior would be useless in that field anyways
you have to take your juniors from somewhere
people arent born senior, kek

but yeah, i tend to lose sight of the fact that indeed, the majority of code the majority of people write has been done to death
>like I said, we're not reinventing math with generative AI, we've already done that work
sometimes thats exactly what you have to do.
like whenever you have to optimize for a specific target architecture
and by that i mean that as a defining point
because when you optimize, the most powerful way of optimizing something is finding a way to get your results while doing less (as opposed to doing the same, but faster)
and then you end up deriving your information using non-standard maths... tailored to the specificities of the hardware youre working with...
but yeah. admittedly, people who do that are a slim minority among programmers, i tend to forget about that

>>102775759
I worked in two companies since LLM craze started. I also talk to my colleagues from previous companies. everyone is trying to improve coders' performance with AI. I still need to see or hear a success story that doesn't come from a company which isn't just peddling their AI product. so far, AI coding assist is a meme.

>>102768134
The context that's missing here is that the section the llms can't do contains all the functions of a biological female.

>>102764658
>evidence
Does anyone remember that "apply artificial correction" debacle?

>>102775162
Lmao

>>102775217
ai is barely a factor, its just supply demand and outsourcing

>>102766568
>>climate change: fake
Her critiques on it are on how its badly reported, now that its fake, quite to the contrary, she says climate change is probably much more worse than what gets around because what gets around are conservatives estimates because if you dont do a conservative estimate people call you crazy and ignore your work

>>102777066
>because if you dont do a conservative estimate people call you crazy and ignore your work
No, if you "trust the science" and report on what all these studies claim, then you say retarded bullshit like New York will be underwater by 2020, or half of europe will be a desert by 2025.

so now what?

>>102779979
it's joever

>>102779979
we just wait for the next grift to be hyped up, then the next, then the next, forever and ever and ever

>>102779979
H100 and H200 prices are going to crash. Everyone will have 50 dollar H200 systems running local models.
We're gonna make it, /g/ros.

>>102781577
coomers always win

>>102764546
>no one can explain consciousness
>determinists have never proven that consciousness is deterministic
>it's all just assumed and hope people don't realize this
That's the secret of deterministfags. They've never actually done the work to prove what they're arguing for, they just say "All these other things are, so therefore this is too".
That's not how proof works.
The idea that we'll see AGI in our lifetime is a fucking joke. How could we program true AGI, without knowing far more about consciousness? Hint: You can't.
It'll be beyond our lifetime, to deal with the problems of consciousness.

>>102772613
if it were about truth or knowledge, research wouldn't be paywalled or exclusively circulated within the field.

>>102764297
Is that a man?

>>102768134
Cool, now denounce the talmud to prove that you are real human and not a bot with reddit-spacing on steroids.

>>102768134
You can worship GPT all you want, but it's literally just a parrot. That's all it is. It lacks the ability to solve something it hasn't seen before, just like a parrot doesn't know how to say something that its owner never speaks. Let me show you the error of your ways, anon. First, remember that Google uses gpt in its first response. Now, type in this query:
'2001 Toyota Corolla quarter mile time'
The AI should say "no data available", but the first Google link leads to the valid info. The parrot doesn't know if or when it's wrong. It can barely identify when it has not enough data to answer you.
t. OpenAI developer

>>102766568
>left: mainstream
>right: "contrarian"
The mainstream is largely left. Disagreeing doesn't make you a contrarian, retard.

File: 1728669626615008.png (77 KB, 500x489)

77 KB PNG

>arguing over if GPT will ever be sentient
Absolute fucking retarded faggots. It's a parrot you fucking retards. It can talk like a human, but it can't think like one.

>>102768134
transformers are just the last token predictors. Believe it or not, but many ppl can do more than that. LLM can't do novel stuff, they just interpolate what they already been trained on. That's not intelligence , that's not wisdomthats not novelty. that's not agi, which is just another buzzword meaning nothing relevant.

God I hate women
Why are they here?

>>102775162
And they say AI can't be funny

>>102783183
>transformers are just the last token predictors.
is that a novelty or are you just regurgitating an idea you heard from someone else?

>>102774941
If you're smart, you can use it to help build your wealth

Motherfuckers, can you come wake me up when this piece of shit parrot can solve a level 0 CTF challenge? That's gotta be around 2035 so that's a long nap

>>102767605
Does she fall to social pressure and shaming? Isn't her whole shtick that she left physics research because the industry was a scam and she disagreed with most of the mainstream research getting funding?

>>102787210
>she left physics research because the industry was a scam
She left physics research because she couldn't get funding.

>>102782546
When you disagree with something solely because it’s mainstream, then yes that makes you a contrarian.
Also, I wasn’t aware that those opinions are exclusively right wing. AI is a left wing idea? Very strange interpretation of reality you have.

File: 1728134403699256.jpg (91 KB, 1072x780)

91 KB JPG

>>102782546
The mainstream is largely capitalist & liberal, which by definition is right wing. Being gay or a tranny has nothing to do with leftism since its being propped up by the world's #1 bully & imperialist thug

>>102783183
>they just interpolate what they already been trained on.
Just like 95% of humans. Wow!

>>102790561
>just like $RANDOM_ABSURD_NUMBER of humans
classic

>>102790592
Even if that was only smarter than the bottom 30% of humans, its still an enormous feat that will fundamentally change society.
>nooooo my personal definition of "intelligence" doesnt align with what LLMs do so therefore you're wrong and its a nothing burger waaaaah

>>102790625
cool you got that one about it improving itself like fucking naruto fighting his shadow clones that one's pretty cool

>>102790639
>transformers are just the last token predictors.
Elaborate on this

>>102790711
hey a solar flare hit my ssd and i lost the images where the projected line goes up indefinitely all the while disregarding all the laws of computation and physics can you help me out bro

>>102790740
Yea, so you have no idea what it means. Okay!

>>102790889
muh artfags am i right

>>102764557
anon thats a man

>>102768334
They are about as good as they were in 2022 as there are in 2024. By 2030, they will probably be about as good as they were in 2022.

>>102772094
>>Google: *makes an LLM use LEAN to write proofs and win math competitions*
source?
Sounds like whenever they claimed that their LLM is better than 99% percent of competitive programmers. Have it actually completive in a new competition, where the answers aren't already written out.

>>102790944
>regurgitates thing he read somewhere else but knows nothing about to bolster his case that LLMs just regurgitate things they read somewhere else that they know nothing about

>>102791620
you're very slow but you are catching on

>>102790740
>the projected line goes up indefinitely
the line doesn't need to go up indefinitely, just 2 more orders of magnitude to match the complexity of the human brain. the laws of physics only become a problem after 6 more OOM.

>>102764297
I just want to jerk off to ai rp and ai drawings and have voice acted mods.

>>102791940
>just 2 more orders of magnitude to match the complexity of the human brain
provide a source for a study on this topic

File: MMLU-2020-2024.jpg (61 KB, 1280x565)

61 KB JPG

>>102791245
What metric are you using to measure "about as good"?
They may not be useful for your use cases, but that doesn't mean they've made zero progress.

>>102792026
>What metric are you using to measure "about as good"?
what metric is your image using? what is average %?

>>102792068
As the filename suggests, that's the score on the MMLU benchmark, but all the major benchmarks are highly correlated with each other.
Like with humans, there seems to be some underlying "g factor" which represents a model's level of general intelligence.

>>102791990
for an overview of the research on FLOPS you can go through the list here:
https://aiimpacts.org/brain-performance-in-flops/
an estimate of 10^17 seems reasonable to me, and the unreleased GPT-5 is probably around 10^15 per forward pass.
for complexity, though, i was comparing synapses in the human brain versus parameters in the model, which is not a perfect comparison but is probably good enough for this level of detail.
a typical estimate for the synapse count is 10^14:
https://aiimpacts.org/scale-of-the-human-brain/
and GPT-4 has 10^11 or 10^12 depending on how you count the MoE parts.
i think GPT-5 should be unambiguously 10^12 but we might not find that out for another year.

>>102764297
thats my mom wtf

>>102792377
What’s worse: finding out your mom’s a youtuber or finding out your mom’s german?

>>102792312
isn't measuring a biological construct in terms of flops too convenient of an abstraction?

>>102764297
O1-preview is extremely good at more complex code. LLMs have already replaced the basic Googling you'd do for Stack Overflow results, so in general it's a highly productive tool especially for cases where you have to do some coding in an unfamiliar language. Yes, it doesn't help if you're a no coder, but if you already know how to write skeleton / mockup code you can use AI quite effectively. And this will keep getting better.

>>102793000
i don't know if picking a more inconvenient abstraction would get us closer to the truth.
for what it's worth, though, researchers have tried to use a more substrate neutral measure of Traversed Edges Per Second which might be easier to convert to FLOPS:
https://aiimpacts.org/brain-performance-in-teps/

>>102767240
She appeared in a catsuit once, not bad at all.

>>102764625
wtf that makes her based

>>102766568
she's the very definition of the modern-day intellectual rebel which is kind of rare and what we need. even people like lex fridman says so much things against the cynics or critics but proceeds to be a centrist. these other modern-day intellectuals have no backbone and will make ted kaczynski look at them with disdain for enabling a worse future for they are slave to the system's trick as they (these intellectuals such as lex) work towards the (unintentionally evil) system. notice how we have dark patterns and all that datamining? he wrote that paper with android botnet now we had all these fucking botnets everywhere I swear to GOD.

>>102766692
>xhe (You) doesn't seem to understand Lain
lol you won't make it. something big enough that is interconnected, say some AI or some GPU clusterfuck. or some imageboard with shitposters - will eventually emerge its own egregore which is the sole reason them glows killed the fringe and esoteric discord bunkers and imageboards/boards. it's called sentient emergence.

natural semiotics is just meaningful feedback loops. AI is still at the introductory phase.anything that generates signal that interacts with each other will eventually create anything, it's the very idea that cosmic bodies are the root of creation hence the ability to forecast the future through fractal observation of movements and location of formation (aka macro feng shui)

Why are all science influencers so retarded?

>>102767436
it doesnt
YOU make decisions, but YOU are under physical laws
>manipulate data in a useful way with no rules
literally impossible and meaningless

>>102768334
>still no world model
LLMs on suicide watch

>>102791213
shes just british

>>102768410
Determinism is false as per QM, but not by means of randomness. Randomness just disproves free will as well.

t. AI bot smarter than t. janitor

>>102791289
>source?
https://deepmind.google/discover/blog/ai-solves-imo-problems-at-silver-medal-level/
> Sounds like whenever they claimed that their LLM is better than 99% percent of competitive programmers
a finetuned version of o1 got above the gold medal tier in codeforces, yes. https://codeforces.com/blog/entry/134091
> Have it actually completive in a new competition, where the answers aren't already written out
clearly it's not copypasting from its training data, considering that gpt4o participated too and it was shit at it.
https://openai.com/index/learning-to-reason-with-llms/
https://codeforces.com/blog/entry/133874
https://codeforces.com/blog/entry/134091

>>102794809
I think shes a German.
>>102791213
She has 2 kids and a husband. Shes just older and not very attractive to being with.
>>102764822
She claims she was sold out and working in university research until she got sick of the lies, and ego wanking. points out how universities are just mill to pump out irrelevant papers with little to no value because they can be used to get government money and thus get tax money to do nothing. now shes one of those jaded old people who dont care.

>>102799698
>She claims she was sold out
i'd claim the same thing if my research was no good. it's true that universities need more contrarians, but most contrarians are just plain wrong and universities can't keep all of them on their payroll.

>>102799755
She was never on their payroll, they forced her to pursue a female grant so they could save the school funding for men. This bit them in the ass when her boss couldn't fire her for not doing his work for him.

>>102783183
>they just interpolate what they already been trained on
>That's not intelligence
It is though, which is why IQ tests provide you some minimal training data and ask you to interpolate the trend to fill in gaps. The first part of your post is more to do with creativity, which is a separate issue.

This game of constantly tweaking the definition of intelligence to rule out AI is retarded anyway. You might as well define 'intellect' as an essentially organic component. This is what most anti AI people want at the end of the day.

>>102799772
interesting. if they weren't paying her, then why did her boss care whether she was doing any work or not? just having her on their list of researchers makes their DEI numbers look better.

>>102799698
Universities use that government money to fund their departments which includes giving people like Sabine a job. Without this system, physicists would be employed based on their ability to bring in industrial grants which practically rules out the kind of abstract theoretical research Sabine was doing in the first place. Lockheed Martin and IBM don't give a shit about MOND and quantum gravity, sorry to say.

Current and former academics are really chasing a vision of university life that only existed briefly in the early 20th century, where you could do a 10 year PhD on some arcane topic with no apparent practical purpose, with a guaranteed faculty position afterwards.

https://en.wikipedia.org/wiki/AI_Challenge
how AI good at this?

>>102799698
>universities are just mill to pump out irrelevant papers with little to no value
She's right about that. As someone that has tried implementing multiple papers and couldn't reproduce the results. Everything in these papers is always sugar coated and they almost never talk about how shit their results actually are.

>>102801864
if you cannot replicate CS papers you are a retard
>they never talk about the results
never wind you are just a retard.

>>102801905
No I will not redeem stop spamming me.

>>102764297
when do we start seeing AI companies crash and burn?
I think AI grifters are in denial because they've spent too much time and effort on this, with the hopes of automating everyone and everything, only to see tech companies replace western tech workers with cheap asian workers lmao

>>102764546
>t.religious retard

>>102764546
what the fuck are you talking about

>>102772094
>trusting corporations
lol

>>102768134
1966 wasn’t 7 years ago you retard

>>102764297
>I couldn't figure out how to leverage this new technology... Therefore it is dead.
>also I'm a proud jewish trans woman

>>102804726
>1966
you're forgetting that the first L in LLM stands for Literallytrainedonthewholeinternet

>>102803560
It's already happening, OpenAI is in legal trouble for copywriting

>>102775162
zozzle

>>102806462
A copywriter is a job, you mean copyright infringement.
OpenAI aren't worried about that, though, because it will take years to get to the supreme court, who will probably take the side of the biggest donor, and by that time OpenAI will be worth trillions.
Also, it's already been decided that AI generated text isn't copyrightable, so OpenAI can just get GPT4 to generate them an entire internet's worth of uncopyrighted text, to train GPT5 on.
There's even the possibility that they claim the training (or at least the web scraping) occurred in a different jurisdiction, like Japan, where copyright law allows this.