If you are trying to learn AI/ML or advanced concepts in computers. Read these

[Jason Voorhees]

Read these books. They are expensive but are a gold mine. They are O'Riley publications. Written by people who have pioneers of these fields and have decades of experience. At the end of every chapter you'll find exercises and problems that you can work out and its not just random easy stuff that you can chatgpt. The questions are legitimately difficult require a lot of googling, reading research papers and understanding concepts on a fundamental level.


Highly recommended th quality of their new books has a declined a lot but their old books are extremely good. Do keep in mind that these books aren't for noobs and more for people in the field with some experience who are trying upskill and learn new stuff. If you are a beginner you may find all of their books to be very overwhelming but once you have the basics covered these books mog any courses or whatever.

 

[mcmentalonthemic]

She Groks me till I Algorithm

 

[ihatereddit]

how close are we to AGI/ASI I dont wanna work nigga

 

[Jason Voorhees]

how close are we to AGI/ASI I dont wanna work nigga

These kind of questions is exactly why I don't bother having any kind of intellectual discussion on this forum @imontheloose

 

[mcmentalonthemic]

how close are we to AGI/ASI I dont wanna work nigga

I honestly don't believe AI can scale much more

https://looksmax.org/threads/yes-ai-was-always-destined-to-fail.1695153

 

[imontheloose]

This kind of questions is exactly why I don't bother having any kind of intellectual questions on this forum @imontheloose

this forum consists of children and retards, often a blend of the 2. quoted user is a known troll

 

[Jason Voorhees]

@topology @Swarthy Knight @Menas @Heretic

 

[mcmentalonthemic]

this forum consists of children and retards, often a blend of the 2. quoted user is a known troll

JFLLL SURE BRO.
You're telling me the guy with the soyjak pfp isn't actually a 145+ IQ ASPD mogger?

 

[Jason Voorhees]

@BigBallsLarry

 

[ihatereddit]

I honestly don't believe AI can scale much more

https://looksmax.org/threads/yes-ai-was-always-destined-to-fail.1695153

Yes the "scaling is all you need" hype died in 2023

These kind of questions is exactly why I don't bother having any kind of intellectual discussion on this forum @imontheloose

I'm trying to think of a less retarded question but Im low IQ tbh

 

[Jason Voorhees]

@JohnDoe since you want to learn devops. Read their books

 

[Jason Voorhees]

@davidlaidisme67 @ltnbrownacnecel @Nexom @adeeyeah @jeoyw9192

 

[Elegant]

will absolutely do this throughout the summer if i am done with all exams (do u think this is still overwhelming for someone still in uni)?

 

[Jason Voorhees]

will absolutely do this throughout the summer if i am done with all exams (do u think this is still overwhelming for someone still in uni)?

Depends on which book you pick up but in general if you are done with the course material covering OS networking, have a couple projects or internships under your belt in that domain then yes more than doable, and potentially super valuable.

 

[Veridic]

Read these books. They are expensive but are a gold mine. They are O'Riley publications. Written by people who have pioneers of these fields and have decades of experience. At the end of every chapter you'll find exercises and problems that you can work out and its not just random easy stuff that you can chatgpt. The questions are legitimately difficult require a lot of googling, reading research papers and understanding concepts on a fundamental level.


Highly recommended th quality of their new books has a declined a lot but their old books are extremely good. Do keep in mind that these books aren't for noobs and more for people in the field with some experience who are trying upskill and learn new stuff. If you are a beginner you may find all of their books to be very overwhelming but once you have the basics covered these books mog any courses or whatever.


View attachment 4650315View attachment 4650316View attachment 4650317View attachment 4650323View attachment 4650324

Super interesting, I will be checking these out however I feel like there’s a bit of a conflation with the quality of the type of questions people ask ai since it’s trained to mirror the users input.

You don’t think that someone with better articulation could utilize ai to help them understand certain concepts?

Low iq inputs will always generate low iq output

 

[widdi]

how close are we to AGI/ASI I dont wanna work nigga

I honestly don't believe AI can scale much more

https://looksmax.org/threads/yes-ai-was-always-destined-to-fail.1695153

AGI already exists. The first principal component explains similar variance (44% for humans, 40-50% for AI), and performance correlates across benchmarks just as it does across human cognitive tests.

Depending on the model and the benchmark it either performs as well as the average human in much less time or even better.

On offline progressive matrices tests it’s scoring between 110 and 130.

It’s also not a bubble at least at the top. P/E ratios are about half as bad as the dot com bubble and revenue is growing faster than valuation for OpenAI, Anthropic, and xAI. The bubble part is from people selling glorified GPT wrappers as part of some SaaS thing.

But I think there are structural limits to it ever being adopted because humans won’t trust it. We’re such a credential driven society that this would be a major barrier. But realistically, we’re already at a place where we could cut down the white collar workforce by 80% and be okay, I just don’t think it will ever happen but that’s not because of any problem with AI’s competence in of itself.

 

[mcmentalonthemic]

AGI already exists.

And then I stopped reading

 

[widdi]

this forum consists of children and retards, often a blend of the 2. quoted user is a known troll

ihatereddit is a comedic genius ur ngmi

 

[Jason Voorhees]

AGI already exists. The first principal component explains similar variance (44% for humans, 40-50% for AI), and performance correlates across benchmarks just as it does across human cognitive tests.

Depending on the model and the benchmark it either performs as well as the average human in much less time or even better.

On offline progressive matrices tests it’s scoring between 110 and 130.

It’s also not a bubble at least at the top. P/E ratios are about half as bad as the dot com bubble and revenue is growing faster than valuation for OpenAI, Anthropic, and xAI. The bubble part is from people selling glorified GPT wrappers as part of some SaaS thing.

But I think there are structural limits to it ever being adopted because humans won’t trust it. We’re such a credential driven society that this would be a major barrier. But realistically, we’re already at a place where we could cut down the white collar workforce by 80% and be okay, I just don’t think it will ever happen but that’s not because of any problem with AI’s competence in of itself.

If you scrolled a little more in that blog you must have read that the the author himself admitted that the 40-50% variance claim is from narrow, saturated benchmarks that doesn't accurately match human g factor's predictive power

 

[messi1]

These kind of questions is exactly why I don't bother having any kind of intellectual discussion on this forum

Most people here aren't cs student

 

[mcmentalonthemic]

Most people here aren't cs student

Neither am I yet I can still have a discussion about this bud, jfl

 

[widdi]

If you scrolled a little more in that blog you must have read that the the author himself admitted that the 40-50% variance claim is from narrow, saturated benchmarks that doesn't accurately match human g factor's predictive power

Where? It’s performing 98% as well as human STEM grads

If you want to say that its level of phenomenology is lower at any IQ score, maybe, but that’s irrelevant for working.

 

[Jason Voorhees]

Where? It’s performing 98% as well as human STEM grads

If you want to say that its level of phenomenology is lower at any IQ score, maybe, but that’s irrelevant for working.

Frontier models only crush some saturated undergrad/grad benchmarks like that MMLU thing that went viral on truly hard STEM stuff like that HLE exam they still be hitting 20-38% accuracy vs human PhD experts at 65%+. That 98% thing only went viral .They're only consistent on routine grad problems but always flop on novel reasoning. I talk to people building these models and they all say tell me we're nowhere near there yet. The gap on hard original work is still large.

 

[JohnDoe]

Pointless learning or going into tech now, all going to be replaced by A.I.

 

[Jason Voorhees]

Pointless learning or going into tech now, all going to be replaced by A.I.

You niggas are a lost cause tbh.

 

[JohnDoe]

You niggas are a lost cause tbh.

Just be honest, we know it's over. A.I will continue getting better and all these niggas who are doing tech now in uni are retarded and will be working at walmart when they graduate.

 

[widdi]

Frontier models only crush some saturated undergrad/grad benchmarks like that MMLU or on truly hard STEM stuff like that HLE exam they still be hitting 20-38% accuracy vs human PhD experts at 65%+. That 98% thing only went viral .They're only consistent on routine grad problems but always flop on novel reasoning. I talk to people building these models and they all say tell me we're nowhere near there yet. The gap on hard original work is still large.

You’re conflating two different measures

“Trained on the internet and almost all published papers and books, even last year’s o1 scored ~80% on GPQA (PhD-level science questions). Gemini 3 Pro now scores 93%. This is superhuman: PhD experts answer 65% correctly, non-experts 34% *despite* web access.”

It’s beating PHD experts already. HLE hasn’t been taken by any human because it’s literally impossible to. Answering 2500 super high level questions would take a decade

In extremely novel reasoning, I agree with you. I don’t think AI will be able to invent something entirely new or create Shakespeare. But, most basic research isn’t even done by corporations, it’s done by the govt and most writing jobs online involve writing random slop.

 

[Jason Voorhees]

Just be honest, we know it's over. A.I will continue getting better and all these niggas who are doing tech now in uni are retarded and will be working at walmart when they graduate.

.I am okay with AI taking jobs as long as it takes everyone out. You want to take my job. Come take it AI do it sooner. Take my managers too. We'll all be in the same boat sailing towards the nearest thela to start selling chai. Imagine all of us jobless tech niggas laying bricks on a construction singing jimmmy Jimmy aaja aaja while the cement mixer goes thak thak thak in rhythm. I'll take it



@Lightskin Ethnic @imontheloose

 

[imontheloose]

.I am okay with AI taking jobs as long as it takes everyone out. You want to take my job. Come take it AI do it sooner. Take my managers too. We'll all be in the same boat sailing towards the nearest thela to start selling chai. Imagine all of us jobless tech niggas laying bricks on a construction singing Sajan ki aankhon mein pyaar dekha pehli baar while the cement mixer goes thak thak thak in rhythm. I'll take it



@Lightskin Ethnic @imontheloose

i dont like them elaichi cardamoms in my chai so all i ask is u dont add it to mine

@JohnDoe

 

[JohnDoe]

.I am okay with AI taking jobs as long as it takes everyone out. You want to take my job. Come take it AI do it sooner. Take my managers too. We'll all be in the same boat sailing towards the nearest thela to start selling chai. Imagine all of us jobless tech niggas laying bricks on a construction singing jimmmy Jimmy aaja aaja while the cement mixer goes thak thak thak in rhythm. I'll take it



@Lightskin Ethnic @imontheloose

A.I is cancer, wish it was never invented

 

[JohnDoe]

i dont like them elaichi cardamoms in my chai so all i ask is u dont add it to mine

@JohnDoe

I have never drank tea, coffee or hot chocolate in my entire life, not a sip. Don't ask for details. I'd rather stand in front of a double decker bus than serve people at starbucks or some south asian chaii place

 

[imontheloose]

I have never drank tea, coffee or hot chocolate in my entire life, not a sip. Don't ask for details. I'd rather stand in front of a double decker bus than serve people at starbucks or some south asian chaii place

do u have autism? /srs

 

[widdi]

Just be honest, we know it's over. A.I will continue getting better and all these niggas who are doing tech now in uni are retarded and will be working at walmart when they graduate.

Companies don’t optimize for hyper profitability though. That’s libertarian autism. Otherwise you’d see them try to constantly undercut each other in a race to the bottom for prices and labor costs. In reality, if you talk to anyone who sets prices, prices are determined by costs + a markup for profit 99% of the time.

Yeah AI will keep getting better and probably destroy the average human even harder next year, but as long as AI + human is even 10% more productive than AI alone companies are gonna be too pussy to pull the plug.

 

[Jason Voorhees]

You’re conflating two different measures

“Trained on the internet and almost all published papers and books, even last year’s o1 scored ~80% on GPQA (PhD-level science questions). Gemini 3 Pro now scores 93%. This is superhuman: PhD experts answer 65% correctly, non-experts 34% *despite* web access.”

It’s beating PHD experts already. HLE hasn’t been taken by any human because it’s literally impossible to. Answering 2500 super high level questions would take a decade

In extremely novel reasoning, I agree with you. I don’t think AI will be able to invent something entirely new or create Shakespeare. But, most basic research isn’t even done by corporations, it’s done by the govt and most writing jobs online involve writing random slop.

the retrieval and pattern matching. at the PhD level is genuinely superhuman level. If you need a walking encyclopedia then yes there's not a shadow of doubt the model wins but a A model scoring 90+% on a PhD exam where the answer is already known and likely exists in some form in the training corpus is not the same as a PhD performing science.

Even the people who created HLE said that models can sniff out answers they still fail at the m step wise logical chain that a human expert uses to verify if an answer is even physically possible. If Al is just replacing government research slop' and SEO writing then sure

 

[Beastimmung]

Your determination to be as cracked as you can in your field is admirable

 

[Jason Voorhees]

.I am okay with AI taking jobs as long as it takes everyone out. You want to take my job. Come take it AI do it sooner. Take my managers too. We'll all be in the same boat sailing towards the nearest thela to start selling chai. Imagine all of us jobless tech niggas laying bricks on a construction singing jimmmy Jimmy aaja aaja while the cement mixer goes thak thak thak in rhythm. I'll take it



@Lightskin Ethnic @imontheloose

@MiserableMan

 

[Jason Voorhees]

A.I is cancer, wish it was never invented

I personally love AI. I use it everyday.

 

[MiserableMan]

@MiserableMan

I am to low intelligence for threads like this bhai. Tag me in the more casual ones.

 

[Jason Voorhees]

I am to low intelligence for threads like this bhai. Tag me in the more casual ones.

I like this song

 

[widdi]

the retrieval and pattern matching. at the PhD level is genuinely superhuman level. If you need a walking encyclopedia then yes there's not a shadow of doubt the model wins but a A model scoring 90+% on a PhD exam where the answer is already known and likely exists in some form in the training corpus is not the same as a PhD performing science.

Even the people who created HLE said that models can sniff out answers they still fail at the m step wise logical chain that a human expert uses to verify if an answer is even physically possible. If Al is just replacing government research slop' and SEO writing then sure

They do well even without internet access, not that it matters because internet access is available for jobs. If it’s only doing retrieval and pattern matching with internet access that’s on the same level of 98% of humans probably. I don’t know about you, but I’ve never derived a complex equation from first principles or invented something new. That’s kind of a ridiculous standard to thrust onto AI if the question is just whether it’s smart enough to replace entry to mid level white collar workers with managerial oversight.

 

[MiserableMan]

I like this song

Yeah it's a nice song, have a nice valentine tomorrow brah.

 

[jeoyw9192]

Read these books. They are expensive but are a gold mine. They are O'Riley publications. Written by people who have pioneers of these fields and have decades of experience. At the end of every chapter you'll find exercises and problems that you can work out and its not just random easy stuff that you can chatgpt. The questions are legitimately difficult require a lot of googling, reading research papers and understanding concepts on a fundamental level.


Highly recommended th quality of their new books has a declined a lot but their old books are extremely good. Do keep in mind that these books aren't for noobs and more for people in the field with some experience who are trying upskill and learn new stuff. If you are a beginner you may find all of their books to be very overwhelming but once you have the basics covered these books mog any courses or whatever.


View attachment 4650315View attachment 4650316View attachment 4650317View attachment 4650323View attachment 4650324

Would also add this to the list, https://mlsysbook.ai/book/, recent ML systems textbook (sorta an extension of a Harvard course taught by the same author of this textbook). Take a look lmk what you think

 

[Scandicel]

Read these books. They are expensive but are a gold mine. They are O'Riley publications. Written by people who have pioneers of these fields and have decades of experience. At the end of every chapter you'll find exercises and problems that you can work out and its not just random easy stuff that you can chatgpt. The questions are legitimately difficult require a lot of googling, reading research papers and understanding concepts on a fundamental level.


Highly recommended th quality of their new books has a declined a lot but their old books are extremely good. Do keep in mind that these books aren't for noobs and more for people in the field with some experience who are trying upskill and learn new stuff. If you are a beginner you may find all of their books to be very overwhelming but once you have the basics covered these books mog any courses or whatever.


View attachment 4650315View attachment 4650316View attachment 4650317View attachment 4650323View attachment 4650324

bookmarked.

 

[Mavueis]

Read these books. They are expensive but are a gold mine. They are O'Riley publications. Written by people who have pioneers of these fields and have decades of experience. At the end of every chapter you'll find exercises and problems that you can work out and its not just random easy stuff that you can chatgpt. The questions are legitimately difficult require a lot of googling, reading research papers and understanding concepts on a fundamental level.


Highly recommended th quality of their new books has a declined a lot but their old books are extremely good. Do keep in mind that these books aren't for noobs and more for people in the field with some experience who are trying upskill and learn new stuff. If you are a beginner you may find all of their books to be very overwhelming but once you have the basics covered these books mog any courses or whatever.


View attachment 4650315View attachment 4650316View attachment 4650317View attachment 4650323View attachment 4650324

bookmarked