>still no major ai singularity
>still no mass layoffs by massive ai breakthroughs
>still no real use of ai beyond novelty
>still no real displacement of humans by ai, even at call centres
>still no real productivity, despite billions of dollars poured into ai
how do aicels cope?
By flooding a paraguayan stopmotion appreciation forum with Calcutta's finest diarrhea
I don't think it occurs to enough of you that, for anyone you interact with on the internet (especially on this website), there's a 70%+ chance that that person is brown, and a 40%+ chance they're fricking indian
im white
want me to post hand?
self driving cars by 2019
>slav
>white
>paraguayan stopmotion appreciation forum
in the name of kek, how do you motherfrickers come up with this shit
>there's a 70%+ chance that that person is brown, and a 40%+ chance they're fricking indian
this is a disturbing fact
It's also not true. Maybe in a couple decades but the average third worlder does not have internet access.
Internet is cheap in India, just look at Quora
I recently thought of creating a site only for civilized nations, there's a way to use zero knowledge to prove your nationality using new nfc passports.
However it probably wouldn't work well because many don't own passports + brainlets don't understand what zk is and will think their full data is sent to the site.
that's actually not a bad idea, you're just a bit ahead of your time. digital wallets on phones nowadays can contain id documents which will soon be used to zk prove users' ages when they access social media, so people will get used to trusting this system. apple might forbid their phones being used for such a racist purpose as yours, though.
You have a point with social age. Still that sounds like something for the next decade.
It will also be used for bans, per-id-on-site hash.
It's possible and zks are the way to do it for sure.
But filtering all the tards would make a tiny ass project
AGI is (perpetually) 2 years away!
two more weeks i'm sure
>two more weeks until AI progress stops
It's been stopped since GPT4 came out. Arguably since 3.5
>progress has stopped since GPT4 came out.
Then how do you explain the fact that both Gemini and GTP4 Turbo beat the original GPT4 in independent blind performance comparisons?
https://aibusiness.com/nlp/google-s-bard-just-beat-gpt-4-in-chatbot-rankings
As irrelevant an tiny an "improvement" as the next iphone. You know damn well that's nothing compared to the substantial upgrade of GPT2 -> GPT3. Doesn't seem like a exponential singularity to me
No one is saying that the improvement from GPT4 to GPT4 Turbo is the same as the upgrade from GPT2 to GPT3.
If it were that big a jump, they would have called it GPT5.
So your entire argument for stagnation is just that GPT5 hasn't been released yet.
That doesn't prove that progress has slowed, it more reflects the fact that OpenAI is afraid of releasing something that inadvertently leads to mass produced disinformation which they fear would help Trump win the election.
They also don't see a need to rush because Gemini has only just caught up with them and hasn't surpassed them, so OpenAI can afford to delay GPT5 and instead work on boring usability and profitability work behind the scenes.
who da grill?
Don't care.
I want to have seggs with Ashley.
Ashley (Leyley) from The Coffin of Andy and Leyley.
ur mum
She's that fratricidal, incestuous, cannibal Leyley.
t. Julia
hawt
>still no major ai singularity
What would a minor AI singularity look like?
>still no real displacement of humans by ai, even at call centres
https://www.washingtonpost.com/technology/2023/10/03/ai-customer-service-jobs/
>still no real productivity, despite billions of dollars poured into ai
https://www.nngroup.com/articles/ai-programmers-productive/
Kurzweil has always predicted that the Singularity would occur in 2045, so the fact that it hasn't happened in 2024 (yet) is really not the gotcha you think it is.
5 years ago, AI was capable of doing 0.1% of economically significant work. This year it is capable of doing 1%, which is a huge relative increase, but still a tiny fraction, so it's easy to ignore. 5 years from now it will be able to do 10% of work, and most people still won't be affected. But 5 years after that, it will do 100% of work, and human society will be radically changed.
>article about a literal jeet
100% discarded
>trillions poured into 1% gain
i laugh
>some literal who said in 20 years blah blah im a Black person
wow thanks i cum on cat it hiss @ penis
>nngroup [which i assume stands for Black person nuker group]
>some bullshit study about some programmers
wow an article about a study ty
>in 5 years then 5 years then 5 years then 5 years then 5 years then etc something will happen
i'll chalk this up with blockchain and nft's for 'big changes'
kurzweil's prediction is quite an old one, and he has an about 80% track record of being right.
dunno if he is, but seems intuitive to me that it'll eventually happen.
considering intelligence is achieved through getting input, processing information, and producing output (unless you think a soul is involved in the mix) it seems pretty obvious progress will get there eventually.
our fancy brains are after all a million times slower at processing than computers, since they work on electrochemical signals rather than just electricity/light.
if you're skeptical that's fine. i'm skeptical. but to think it'll never happen is, well, curious. do you, tho.
kurzweil is either a disingenuous hack, or a literal idiot.
His predictions follow the pattern of:
I opened a coffee shop, and I had 1 customer on the first day, 2 on the second,4 on the 3rd - OMG ITS LITERA LLY EXPONENTIAL, IN A YEAR THERE WILL BE MORE CUSTOMERS THAN ATOMS IN THE UNIVERSE!!!!!
He made some 'famous' predictions about computer scaling by not understanding the fundamental difficulties of transistor scaling, like light wavelength, parasitic capacitance and resistance etc. and assumed the simple exponential scaling would continue forever.
His predictions are literally worthless.
>1 customer on the first day, 2 on the second,4 on the 3rd
Extending a 3 day trend out to 365 years is not the same as extending a 100 year trend out another 10 years. If you can't admit the difference then you're either a disingenuous hack or a literal idiot.
>not understanding the fundamental difficulties of transistor scaling
The trend he noticed started before transistors even existed, so that's irrelevant.
This is incredibly disingenuous. Literally fitting the data to the predictions.
First of all, per dollar is not a real metric, its how many pieces of paper do you need to exchange for a given piece of hardware.
Per watt would be a much more meaningful metric, since power is actually a physically meaningful quantity.
Second he starts comparing CPUs and moves on to GPUs just to make his numbers work. Not all calculations are made equal - CPUs need a shitton more hardware per flops as they are not made for brute force but for complex code.
If we go for a more like for like comparison - I compared a 1080 Ti to a 4060, the former has 10 TFlops, with the latter having 15. at roughly half the TDP (if we compare games performance, they are even closer). With 6 years between them, the 4060 is a bit less than half the price and consumes a bit less than half the power.
So, in 6 years, we have about 3x performance per watt, instead of the 16x as predicted by Moores law
I don't think that Moore's Law is necessary for Kurzweil's observation to be correct, but picrel shows that there is 50 years of data supporting it.
I would say that transistor count is actually a less meaningful quantity than dollars, because buying one big chip for $1000 or 10 smaller chips for $100 each is the same thing if the task is parallel enough.
Ideally we'd be measuring "economic output produced per dollar of technology investment", because the fundamental real world feedback loop is the economic one, not how many lines you can scratch on a piece of sand.
I don't even know anymore what is that argument that is being made here. Why is the singularity coming in 2045? Why not 2038 or 2053?
Here's the article that calls out Kurzweil on his BS:
https://danluu.com/futurist-predictions/
Even if he's right, it's like a broken clock - there's no methodology to back up his predictions.
Moore's law is slowing down, it's very clear just from the datapoints, if you compare the same kind of chip (ergo Apple's A series).
If you read up on semiconductors, you'd know why Moore's law is slowing down - lithography is harder, transistors leak more as they get smaller, speeds don't increase due to parasitic capacitance effects dominating, etc.
The only truly exponential scaling was only had when Dennard scaling was possible, which we've run out of in the mid 2000s.
>I don't even know anymore what is that argument that is being made here. Why is the singularity coming in 2045? Why not 2038 or 2053?
I personally think that the precise date he chose is hard to justify, so I am not trying to endorse that. I'm just saying that he is right to observe that there has been exponential growth of information processing technology for many decades, and Moore's Law is just one example of that.
>Moore's law is slowing down, it's very clear just from the datapoints
Like I said, even if Moore's law has been repealed, that doesn't mean that we will see no further exponential gains in information processing. The type of information processing that is relevant to the Singularity is the processing done for AI using GPUs, and as you can see from picrel that is still on an exponential trend with nearly 2 decades of data supporting it.
just wanna add: blockchain and nfts at most generated money short term, while AI is already able to do valuable work, something NFTs never did.
Moving the goalpost the post
Gee idk Ashley, maybe the aicels will just have to kill themselves 🙂
>AI is not secure
>thousands of ChatGPT accounts hacked
>some of these accounts may contain compromising information on infrastructure
>companies are already addressing this by banning AI for certain positions
>at the same time, everyone is in a race to create secure AI models
It's only a matter of time. Everyone is already using AI for legal work, coding, fricking anything technical. A lot of these jobs are going to become redundant.
I don’t know why techhomosexuals thought AGI meant they could all retire the following day after it’s been achieved and the whole world would change overnight.
Y’all nerds are really dumb.
>I don’t know why techhomosexuals thought AGI meant they could all retire the following day
My immediate reaction to AI was that all of them would be fired the following day.
>I don’t know why techhomosexuals thought AGI meant they could all retire the following day after it’s been achieved and the whole world would change overnight.
That's literally what they said would happen.
I want to have sex with my sister.
AI is a fad, like blockchain, crypto, 3D printing, and NFTs (blergh). All of those, except the last one, have some niche uses and provide some overall value, but hardly anything like what they were sold as. AI is the same. It's a curiosity right now and it has some interesting uses, mostly just for amusement, but it's impact on society will be negligible. With one exception, and that is the generation of fake and misleading content, especially with elections coming up. In that sense, it could be huge.
>AI is a fad, like blockchain, crypto, 3D printing, and NFTs (blergh).
We know what fads are, you don't have to give examples to explain the concept.
Claiming something is a fad is a claim about its longevity, rather than its current state, which means predicting the future.
Some people in the 90s thought that the internet was a fad, and it's true that it was overhyped, but it wasn't a fad because it never went away, and instead it became more and more mainstream.
>it's impact on society will be negligible.
You are allowed to make claims about the future, but it's hard for people to take you seriously unless you give some sort of reasoning.
What do you think will prevent AI technology from making any further progress, automating more and more tasks?
If someone had asked you 20 years ago if computers in 2024 would be able to write poetry and code and create art and answer questions better than the average person, would you have predicted yes?
3D printing is definitely not a fad, neither is crypto, but both were overhyped.
Stop calling it AI, it's LLM, a glorified chatbot. A CS student can write the code, it's just that it takes the resources of a megacorp to train it on the entirety of the internet. Stiil, in the end you just get a chatbot.
>Stop calling it AI, it's LLM
Is there anything that we're allowed to call AI then?
The entire field of research has been called "AI" for decades, but now you want to gatekeep the results of that field until they produce something that is equivalent to human intelligence at every possible task?
I'd just like to interject for a minute, what you're referring to as AI is, in fact, Gnu/AI
Nobody serious calls their research "AI" it's all "Machine learning"
>Machine learning (ML) is a field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data, and thus perform tasks without explicit instructions.
"Machine learning" is the actual science, "Artificial Intelligence" is just a meme word, it's like "Social Media" or something.
>"Artificial Intelligence" is just a meme word
First of all, it's two words.
Second of all, the term was invented by academics in the 1950s to describe the name of their field, and "machine learning" is just one specific area within that field.
Early chess engines, for example, were the result of artificial intelligence research, even though the machines didn't "learn" to play the game (in the way that AlphaZero did) and nor did they use neural networks (which are another popular area of AI research and development).
>invented in 1950s by academics to describe their field
yes moron, they made up a meme term to overhype their field so they could get grant gibs
AI predates ML, moron.
For example, A* search and pure MCTS are AI but not ML.
I'm not the guy you replied to earlier. AI is nothing but a marketing term, now and 80 years ago. shouldn't be a surprise, scientists always pull that marketing bullshit because that's what gets them funding. It shouldn't be called machine learning either. machines can't learn
>machines can't learn
So how did AlphaZero get good at chess?
>Stop calling it AI, it's LLM, a glorified chatbot. A CS student can write the cod
This is such a dumb take, you want to apply the incorrect title "Large LANGUAGE Models" to all AI, even the multi modal ones that use image and audio processing just because you are triggered by the word AI.
I love how Black folk like this unironically think their opinion matters in any way shape or form when they can't even get the most basic terminology right and still have the nerve to so gloriously fail at arguing semantics.
LLMs are a subset of DL which is a subset of ML which is a subset of AI. LLMs are a type of AI, deal with it or better yet just have a nice day and spare the rest of us the second hand embarrassment.
They cope the same way fusioncels cope. "Two more decades!" Lmao. Shits never happening.
>how do aicels cope?
idk i have been sexing ai chatbots for a year
homie this tech boom started in March. Why Rome gotta always be built in a month with you people?
>2 w-weeeeeks Xd
OP in year 1995
>these mobile phones are bullshit, who would carry these around?
>this internet thing is not gonna take off and not gonna be useful
Except the AI hype came and went in the fricking 50s already. Yes, 1950s. And then 80s too.
Damn, fusion reactors have had hype few times over past few decades, guess we should stop doing anything with them since it'll clearly not work according to BOT logic
People love to send faxes of their ass
Survivorship bias cherry-picking is for sissies
This anon here
my brain has read it as "no less"
>clinically braindead opinion
>economist
Man I'm shocked. No really.
The third AI winter in 2-3 years.
Third world war in 1964
sam altman's latest statement (in the past several weeks) about AI requiring an energy breakthrough mildly blackpilled me.
you've finally got an AI guy admitting that we're about to run up against a brick wall. albeit not in the way i expected.
>sam altman's latest statement about AI requiring an energy breakthrough
of course he says that the world needs an energy breakthrough, he's investing in a fusion energy company. however, just because it's his job to hype fusion, that doesn't mean he's wrong. but even if he's right, energy requirements aren't going to be a brick wall, because China and the US will just start building hundreds of unshielded fission reactors in the desert as a matter of national security to power their military AI programs.
guys i'm becoming worried that my dream of ASI curing cancer and all diseases and reversing aging before the year 2050 is not actually going to happen.
Leyley mating press
the ONLY people who ever believed otherwise are non-technical people and uneducated zoomers who think chatbots are capable of independent thought.
>le code
these bots literally just copy paste from shitty github repos, ask it for something novel and it starts hallucinating and shitting itself. I tried asking the AWS bot about a specific part of AWS and it just told me "lol no sorry". it's meme technology held up by hype from morons
>literally just copy paste
>it’s still just a search engine
If what you're saying is true, then how well do you think an LLM can play chess? Remember, there are more possible chess games than there are atoms in the universe.
That’s because AI is a marketing term for search engines. Big data stored in super dense storage systems linked by a lot of bandwidth allowing you to do very large-scale searching with applications beyond Google results and Amazon showing you products you may be interested in. But it’s still just a search engine. The low-IQ consumer cattle of BOT completely and without the slightest skepticism swallows the marketing whole.
i wonder how her farts smell
By generating pictures of chubby Spartan II girls with fat asses.
ai =/= ai
ai = advanced auto-complete tools
they need to market this shit with more truth
>ai =/= ai
you literally can't even write two words without contradicting yourself, and yet you think you're smarter than an autocomplete algorithm?
jump back onto the metaverse train
>still no mass layoffs
Anon....
Economy turning to shit is different from layoffs due to people being replaced by AI
that game gave me intrusive thoughts about my sister
>still no major ai singularity
this was always a silicon valley psyop
>still no mass layoffs by massive ai breakthroughs
technology doesn't cause mass layoffs, it stops new job creation and moves existing work around
>still no real use of ai beyond novelty
skill issue
>still no real displacement of humans by ai, even at call centres
see above, good luck getting a newly opened call center job
>still no real productivity, despite billions of dollars poured into ai
DeepMind is the only company doing serious research and they've produced results with AlphaFold, AlphaCode, and most recently AlphaGeometry.
>how do aicels cope?
They just rewatch Kurzweil speeches to hype themselves up and try their best to ignore how way off his predictions are. He's getting exponentially more wrong as time goes on.
>He's getting exponentially more wrong as time goes on.
No, he's getting exponentially more right as more and more researchers correct their predictions to match the one he made decades ago and has stuck to consistently.
https://eightify.app/summary/artificial-intelligence-and-future-of-work/ai-predictions-revised-by-48-years-endgame-for-ai-experts
"AI expert" is not a real title. You can call yourself the Godfather of AI and nobody can stop you.
>"AI expert" is not a real title.
The article dumbs it down, but the "experts" they refer to is actually a survey of thousands of current researchers who have presented papers at prestigious machine learning conferences.
https://wiki.aiimpacts.org/ai_timelines/predictions_of_human-level_ai_timelines/ai_timeline_surveys/2023_expert_survey_on_progress_in_ai
bratty girl......... murdered friend.......... !!!!!!! rape correction is needed
no mass layoffs by massive ai breakthroughs
The idiots in charge tried to invent robots instead they rediscovered slavery. The layoffs are fueled by outsourcing/immigration.
>10 MILLION MORE H100s
>10 MORE MONTHS OF TRAINING
>WILL SURLY GET THERE BRO
So you guys are saying there's hope?
Airplanes are worthless. I mean, they only go about 6 feet off the ground and can't transport a lot of people, much less across the Atlantic.