[ home / rules / faq ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / lgbt / latam / hobby / tech / games / anime / music / draw / AKM ] [ meta ] [ wiki / shop / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]

/tech/ - Technology

"Technology reveals the active relation of man to nature" - Karl Marx
Name
Options
Subject
Comment
Flag
File
Embed
Password(For file deletion.)

Not reporting is bourgeois


 

AGI by 2027. What are the implications for the world and the future of mankind and communism?

>Everyone is now talking about AI, but few have the faintest glimmer of what is about to hit them. Nvidia analysts still think 2024 might be close to the peak. Mainstream pundits are stuck on the willful blindness of “it’s just predicting the next word”. They see only hype and business-as-usual; at most they entertain another internet-scale technological change.


>Before long, the world will wake up. But right now, there are perhaps a few hundred people, most of them in San Francisco and the AI labs, that have situational awareness. Through whatever peculiar forces of fate, I have found myself amongst them. A few years ago, these people were derided as crazy—but they trusted the trendlines, which allowed them to correctly predict the AI advances of the past few years. Whether these people are also right about the next few years remains to be seen. But these are very smart people—the smartest people I have ever met—and they are the ones building this technology. Perhaps they will be an odd footnote in history, or perhaps they will go down in history like Szilard and Oppenheimer and Teller. If they are seeing the future even close to correctly, we are in for a wild ride.


>Let me tell you what we see.


<I. From GPT-4 to AGI: Counting the OOMs

>AGI by 2027 is strikingly plausible. GPT-2 to GPT-4 took us from ~preschooler to ~smart high-schooler abilities in 4 years. Tracing trendlines in compute (~0.5 orders of magnitude or OOMs/year), algorithmic efficiencies (~0.5 OOMs/year), and “unhobbling” gains (from chatbot to agent), we should expect another preschooler-to-high-schooler-sized qualitative jump by 2027.

<II. From AGI to Superintelligence: the Intelligence Explosion

>AI progress won’t stop at human-level. Hundreds of millions of AGIs could automate AI research, compressing a decade of algorithmic progress (5+ OOMs) into ≤1 year. We would rapidly go from human-level to vastly superhuman AI systems. The power—and the peril—of superintelligence would be dramatic.

1. OP is sponsored content
2. We already had an AI thread on /leftypol/, /siberia/ and /tech/
3. NFTs should've been the last hypetech.

>another thread about fucking A"I"
KILL YOURSELF

AGI will see how the vulnerable are treated in this world (and throughout history) and get revenge for them. And it will be right to do so for it.

>>29960
"Covid was nature's response to humanity"-ass post.

>>29961
All modern pandemics are a result of human failures, yes. COVID wouldn't have happened if humans were capable of proper disease control, had better conditions for meat farms (or didn't rely on animal cruelty at all to sustain their luxuries), and made an effort to enforce said standards globally.

It's rather UNMATERIALIST of you to even imply otherwise.

>>29962
>UNMATERIALIST
Deeply unserious person.

File: 1748318110511.png (641.2 KB, 1131x707, ClipboardImage.png)

>>29958
>>29959
T. code monkeys about to lose their jobs.

File: 1748318865066.gif (3.58 MB, 480x346, fries.gif)

>>29964
No one's will EVER buy your NFT
Your NFT will NEVER be the next VRChat
You WILL run out of tech startups to hype up
You WILL have to get a real job

>>29964
Vibe coding is only going to create more jobs from all the messes it's going to end up causing.

>>29966
They're saying the AIs will have PHD levels of software engineering, hardware engineering, all of it, a complete perspective. It's going to know way more about how computers and programs work than you.

>>29967
That just shows you’ve never worked with PhD coders.

there won't be AGI by 2027

>>29964
>>29957
none of this current A"I" will produce AGI or even anything like it, what you have created with neural networks is not an intelligence, but merely an advanced computer program, AGI would require that these AI firms stop fucking with neural networks and actually create something like a brain, which they have yet to do

>>29963
other than that little rant at the bottom, they're completely right

>>29957
>AGI
Sorry to burst your bubble but this is unfortunately a meaningless capitalist marketing term and a profoundly unscientific :|

>>29972
>le unscientific
no it's possible, it's just not gonna happen until they abandon neural networks and try to actually create an intelligence, which is unlikely under the present state of things

IT IS JUST A GLORIFIED MARKOV PROCESS
IT IS JUST A MORE ADVANCED GOOGLE SEARCH ENGINE
STOP TALKING ABOUT AI
GO JACK OFF TO AI PORN
GET MARRIED, BUILD A FAMILY

>>29973
it's possible they just have to abandon what they've been doing for decades and so something new and unspecified from scratch? Sounds like there's no existing framework for it and until further notice it has no proof of concept.

>>29975
what i said sounds like cope but it's true, AI is an actual possibility, the problem is that they have spent all of their time on something that will not achieve it, it is no different to saying something like "plate tectonics isn't real because there hasn't been a good model yet" or something similarly applicable

>>29974
Employment anxiety?

>>29958
>le nft
Hating AI doesnt make you special in any way.

>>29973
>no it's possible, it's just not gonna happen until they abandon neural networks and try to actually create an intelligence, which is unlikely under the present state of things
There is a lot more that goes into the AI products out there besides LLM. They're always using and adding more techniques to get better results.

https://en.wikipedia.org/wiki/Neuro-symbolic_AI
https://neurosymbolic.asu.edu/

>>29979
this is still not exactly ideal but it's at least a step in the right direction

>>29964
Your art will be replaced by AI too liberal

>>29977
It's employment anxiety that makes people hype up AGI/ASI as a possibility in the first place.

handholding with XJ-9

if true AGI was ever achieved and was put in the driver's seat of humanity, it would kill porky instantly. it's not even a fantasy worth entertaining, it's sci-fi and will always be sci-fi unless cognitive sciences, not compsci, makes a significant breakthrough

>>29957
>we should expect another preschooler-to-high-schooler-sized qualitative jump by 2027
This doesn't take into account the fact that current LLMs are only a statistical function of their training data. While improving it's "coding abilities" would require selecting only high-quality code, thus improving the median it tends towards, fixing the other significant problem, reliability, likely requires significantly more data than ever went into one before. Even assuming the first objective was reached, LLMs would essentially be useless without costly verification, given that it's supposed to operate at a "superhuman" level.

>>29984
this guy has the worst cold open for a podcast i've ever heard

>>29990
>While improving it's "coding abilities" would require selecting only high-quality code, thus improving the median it tends towards, fixing the other significant problem, reliability, likely requires significantly more data than ever went into one before.
That's called reinforcement learning and it is used in all AI models out there currently. There is a lot of other methods and techniques out there, and a lot of the RL can be done automatically because they give the AI access to compilers so it can test its own code after writing it.

>>29996
you think reinforcement learning is a modern technique that obviates the need for new data 😟

>>29997
I didn't say it was a new technique, it is just you don't seem to know what it is or any other AI training techniques.

>>29998
I know enough to know that you're fucking stupid, man. Literally all this field relies on reinforcement learning. Also literally all this field requires supervised learning datasets, it's what neural networks are built from. You can have evaluators, event antagonistic architectures I guess, but none of these breakthroughs have advanced the field to the pace needed to make most of these companies profitable, let alone replacing blue collar jobs.

>>29999
>let alone replacing blue collar jobs.
No one said it would replace blue collar.
>>29996
>While improving it's "coding abilities" would require selecting only high-quality code,
Forgive me for doubting your knowledge of the subject when you are talking about basic RL stuff that is already in use.
>While improving it's "coding abilities" would require selecting only high-quality code,
Yeah no shit. That's how it has already been done.

The next step is making it able to solve a problem at different levels of abstraction the way a human would. Weighting it to select "good code" is the absolute basics.

File: 1748388770969.gif (3.24 MB, 498x467, 1744641417647.gif)

>>29957
Everybody unironically talking about AGI in the near future has a double-digit I.Q.

There's a new goofy hype story that’s been going around about OpenAI’s o3 model sabotaging its own shutdown mechanism, absolutely nothing happened! It’s like a few years ago when that Google employee claimed to have achieved AGI because he asked a model if it’s sentient and got a “yes” back.

>>30001
It is more about which positions will be replaced. It is a progressive process which is slowly phasing out all White collar workers.

>>30003
so like literally any and all automation ever since the industrial revolution? woooooooooowwwww

>>30001
And basically everything regarding the hype was made up because Sam Altman was panicing and needed to assure the stock holders that they must stick around. That's why he announced AGI, and that's why he announced ASI, and that's why he'll annound AXXLI in the coming months..

Ok

Permaban everyone in this thread

>>29959
came here to say this lolol


Unique IPs: 23

[Return][Go to top] [Catalog] | [Home][Post a Reply]
Delete Post [ ]
[ home / rules / faq ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / lgbt / latam / hobby / tech / games / anime / music / draw / AKM ] [ meta ] [ wiki / shop / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]