Social War Room Lounge Thread #325: PotWR Edition

Status
Not open for further replies.
Rob Battisti said:
I’m claiming that when super intelligent AGI arrives, it will replace upwards of 80% of jobs done in the world within 5-10 years. Entire industries will cease to exist.

Human work will be obsolete. That’s exactly right.
Click to expand...

I'm not so sure AGI is as close as some people say. Currently what we call "AI" are just LLMs/LRMs, which work by being fed a metric ton of data and analyzing what words commonly appear next to each other so that they can guess the next word in a sequence. They formulate answers by reading a question and then guessing the next word, one by one, until they have a complete response. This is nothing like how human intelligence, or cognition works, and has limitations.

Apple published a paper earlier this year talking about the limitations of LLMs. Basically when certain puzzles reach a threshold of complexity, the effectiveness of the LLMs/LRMs collapses to zero, even when provided the solution to the puzzle.

Probably the assumption for AGI being near is that LLMs/LRMs will not plateau, and simply continue to improve until we somehow get AGI. But the Apple paper is suggesting there's an inherent scaling limit in LLMs, which makes sense if you think about how they work. So it seems like a completely new paradigm may be needed for AGI, let alone super intelligent AGI... and if so, who knows how close or far that is.
 
Rob Battisti said:
The ignorance of it really does blow me away
Click to expand...
Do you really think if your prediction turns out to be accurate people will be unable to find a way to be productive and improve their individual standards of living?

FWIW, regardless of the outcome, I think it's rather further off into the future than you do, however I can see how cumulative advances may bring about a truly general AI much more quickly than can be accurately predicted right now.
 
MeatheadMike said:
I'm not so sure AGI is as close as some people say. Currently what we call "AI" are just LLMs/LRMs, which work by being fed a metric ton of data and analyzing what words commonly appear next to each other so that they can guess the next word in a sequence. They formulate answers by reading a question and then guessing the next word, one by one, until they have a complete response. This is nothing like how human intelligence, or cognition works, and has limitations.

Apple published a paper earlier this year talking about the limitations of LLMs. Basically when certain puzzles reach a threshold of complexity, the effectiveness of the LLMs/LRMs collapses to zero, even when provided the solution to the puzzle.

Probably the assumption for AGI being near is that LLMs/LRMs will not plateau, and simply continue to improve until we somehow get AGI. But the Apple paper is suggesting there's an inherent scaling limit in LLMs, which makes sense if you think about how they work. So it seems like a completely new paradigm may be needed for AGI, let alone super intelligent AGI... and if so, who knows how close or far that is.
Click to expand...
This is an outlier compared the the other leading minds in the space. Apple is also woefully behind in AI.
Andy Capp said:
Do you really think if your prediction turns out to be accurate people will be unable to find a way to be productive and improve their individual standards of living?

FWIW, regardless of the outcome, I think it's rather further off into the future than you do, however I can see how cumulative advances may bring about a truly general AI much more quickly than can be accurately predicted right now.
Click to expand...

I think we have 10 years left. Some leaders say 2030 but that feels wrong.

If it’s 2035 when we achieve AGI and super intelligence, I’d give it just a matter of 2-3 years before it’s completely integrated.

At least your talking about WHEN not IF like others are here.

As for will people remain productive? No. I don’t. Not at scale.

We will need UBI.
 
Rob Battisti said:
This is an outlier compared the the other leading minds in the space. Apple is also woefully behind in AI.


I think we have 10 years left. Some leaders say 2030 but that feels wrong.

If it’s 2035 when we achieve AGI and super intelligence, I’d give it just a matter of 2-3 years before it’s completely integrated.

At least your talking about WHEN not IF like others are here.

As for will people remain productive? No. I don’t. Not at scale.

We will need UBI.
Click to expand...

I don't think Apple themselves being ahead or behind matters; they used state of the art LLMs in their testing (o3-mini, DeepSeek, Claude) to come up with their findings.

Also Apple aren't the only ones who don't believe LLMs alone will yield AGI. Do you?
 
Andy Capp said:
You've been on this forum longer than me and you have to ask? ;)
Click to expand...
True.
Limbo Pete said:
Can you envision a future slightly less fantastic than the plot to Battlestar Galactica?
Click to expand...
I wish it wasn’t going to be like that but it is.

When you can employ 10,000 AI agents who work 24/7 for a fraction of the cost and who are smarter and more capable than someone with a PhD, why would you hire a human at 50/hr who works 40 hours a week and regularly makes mistakes?

The only limit to when ASI is going to come into creation is power.
 
MeatheadMike said:
I don't think Apple themselves being ahead or behind matters, they used state of the art LLMs in their testing (o3-mini, DeepSeek, Claude) to come up with their findings.

Also Apple aren't the only ones who don't believe LLMs alone will yield AGI. Do you?
Click to expand...
I’m less confident that LLMs will be mechanism to ASI but I am confident that LLMs employed agentically will be the key to creating ASI.
 
Rob Battisti said:
True.

I wish it wasn’t going to be like that but it is.

When you can employ 10,000 AI agents who work 24/7 for a fraction of the cost and who are smarter and more capable than someone with a PhD, why would you hire a human at 50/hr who works 40 hours a week and regularly makes mistakes?

The only limit to when ASI is going to come into creation is power.
Click to expand...
 
Rob Battisti said:
I’m less confident that LLMs will be mechanism to ASI but I am confident that LLMs employed agentically will be the key to creating ASI.
Click to expand...

I think that makes sense for ASI. But we would likely still need some separate form of AGI capable of basic human-like intelligence and problem solving to utilize the LLM agents (similar to how we use them now, but more efficiently), and that is what may require an entirely new paradigm, which may not neccesarily be on the horizon.
 
Mr Holmes said:
You still need people to feed it all the different scenarios, analyze the results, apply it appropriately, test it, adjust when it gets things wrong, re-apply, etc. Even knowing what to ask it requires you to have the appropriate background.


I guess we'll see. My estimate is I'll be long retired before we get there.
Click to expand...
One mistake AI enthusiasts make is assuming a continuous, linear growth in the technology's capabilities as well as assuming that it's going to make qualitative leaps.

Cars initially had top speeds of ~10-20mph and yes they've increased a lot since then but we didn't end up with supersonic flying cars. No one today thinks increasing top speeds or designing flying cars is desirable much less inevitable, the focus is on incremental increases to comfort and safety.

We might see something similar with AI where it settles into certain use cases and additional innovations simply better adapt it to those niches rather than seeing quantum leaps in the overall capabilities. AI might take the jobs of graphic designers who make Youtube thumbnails but I don't think they're going to replace Christopher Nolan or Hayao Miyazaki anytime soon.

If anything coding might be the single best use case for modern AI and even there you can see it has a long way to go.
 
Last edited:
Rob Battisti said:
Exactly.

The Industrial Revolution replaced muscle. AI I’ll replace intelligence.

Once again, what is your well thought out take on the future?
Click to expand...
The industrial revolution didn't replace muscle, it made it more efficient. Industrial technologies make construction and agriculture more productive but they still require labor intensive jobs.
 
Islam Imamate said:
The industrial revolution didn't replace muscle, it made it more efficient. Industrial technologies make construction and agriculture more productive but they still require labor intensive jobs.
Click to expand...
Come now, it's obvious Super Grok will make us all obsolete. What use does humanity have when the Supreme Intelligence can beam "pussy in bio" directly to the underside of your eyelids
 
Status
Not open for further replies.

Latest posts

Forum statistics

Threads
1,280,941
Messages
58,329,516
Members
176,000
Latest member
Markooz86

Share this page

Back
Top