Artificial General Intelligence (AGI) News and Discussions

[Cyber_Rebel]

This is concerning. Assuming technology moves forward at exponential pace including something like AGI, if society "changes surprisingly little" that sounds an awful lot like some Cyberpunk style future. I understand the infrastructure and hardware infrastructure needing time to be in place, but people having the same apathy with something capable enough to reshape the economy doesn't sound right to me. Assuming everyday people have access to it, would it not be life changing for that individual's capabilities?

[firestar464]

Average Joe will just go down the "work is sacred" and "ban AI" pipeline

[wjfox]

[firestar464]

Sam says 2025. God I hope he finds some sort of way to keep it out of the hands of Trump and Elon

[wjfox]

[Cyber_Rebel]

For context on who this guy is, he's apparently someone who worked closely with the Google team who created the transformer architecture before working in technical staff at Open AI.

[firestar464]

Not sure they've reached "able to do stuff that an average person can at the level of said average person" though.

[Powers]

Not sure they've reached "able to do stuff that an average person can at the level of said average person" though.

In my opinion being able to do it alone (yeah this again) would be enough, for now.

[spryfusion]

https://arcprize.org/blog/oai-o3-pub-breakthrough

Despite the significant cost per task, these numbers aren't just the result of applying brute force compute to the benchmark. OpenAI's new o3 model represents a significant leap forward in AI's ability to adapt to novel tasks. This is not merely incremental improvement, but a genuine breakthrough, marking a qualitative shift in AI capabilities compared to the prior limitations of LLMs. o3 is a system capable of adapting to tasks it has never encountered before, arguably approaching human-level performance in the ARC-AGI domain.

[Doozer]

If I were to guess, our first AGI-based chatbot will be GPT-6 and come out two years from now.

[Cyber_Rebel]

Dario's timeline of 2026 for "powerful AI" or AGI seems right on the ticket. By that time 03's reasoning cost would have dramatically dropped and it's not too crazy to believe that we'll near some 0-8 kind of reasoner if scaling test time compute moves as swiftly. 3 months is insane if the rate remains the same. If anything, that's a bit beyond "median" human baseline, so it falls to these AIs being as capable with their reasoning in everyday (not just in the lab) tasks. Robotics will move swiftly but there will always be a lag in the physical domain; an agent however, with this kind of ability and assuming cost efficiency really impacts every single knowledge-based sector.

We really need to begin discussing economic frameworks like yesterday. There were never any "walls" or limits and people pretending like there ever was, is going to lead to unemployed outcomes.

[funkervogt]

If I were to guess, our first AGI-based chatbot will be GPT-6 and come out two years from now.

The projections I've seen indicate GPT-6 won't be built before 2029.
https://www.astralcodexten.com/p/sam-al ... 7-trillion
https://epoch.ai/blog/can-ai-scaling-co ... rough-2030

[Doozer]

If I were to guess, our first AGI-based chatbot will be GPT-6 and come out two years from now.

The projections I've seen indicate GPT-6 won't be built before 2029.
https://www.astralcodexten.com/p/sam-al ... 7-trillion
https://epoch.ai/blog/can-ai-scaling-co ... rough-2030

Before then could we at least get a GPT-5.5?

[firestar464]

WSJ- The Next Great Leap in AI Is Behind Schedule and Crazy Expensive

https://archive.ph/L7fOF

It seems that Ilya was right that just feeding data has hit a wall (though people misunderstood this as AI as a whole hitting a wall). Reasoning seems to be the way forward.

[funkervogt]

WSJ- The Next Great Leap in AI Is Behind Schedule and Crazy Expensive

https://archive.ph/L7fOF

It seems that Ilya was right that just feeding data has hit a wall (though people misunderstood this as AI as a whole hitting a wall). Reasoning seems to be the way forward.

I think it's more nuanced. They fed more data in GPT-5 and it was better than GPT-4, but not by as big of a margin as they wanted, so they're still working on it before releasing it. It's taking longer than expected, and it casts into question whether scaling up the data is a feasible way to make GPT-6.

[erowind]

Dario's timeline of 2026 for "powerful AI" or AGI seems right on the ticket. By that time 03's reasoning cost would have dramatically dropped and it's not too crazy to believe that we'll near some 0-8 kind of reasoner if scaling test time compute moves as swiftly. 3 months is insane if the rate remains the same. If anything, that's a bit beyond "median" human baseline, so it falls to these AIs being as capable with their reasoning in everyday (not just in the lab) tasks. Robotics will move swiftly but there will always be a lag in the physical domain; an agent however, with this kind of ability and assuming cost efficiency really impacts every single knowledge-based sector.

We really need to begin discussing economic frameworks like yesterday. There were never any "walls" or limits and people pretending like there ever was, is going to lead to unemployed outcomes.

Cybernetic planning has been possible in some capacity for the past 150 years, in a workable capacity since the 70s, and in a highly advanced utopian manner since the mid 2000s. By the mid 2000s every economic input and output had already been modeled by the leading exchanges and capital firms like blackrock and the NYSE. There came into existence algorithmic trading firms so advanced they scalp fractions of a penny off of market movements they can reliably predict enough to consistently profit over a period of decades across the highs and troughs of the business cycle.

This is all to say that I doubt that enough people in positions of power will be open to any kind of real systemic change until they are faced with the choice of having their heads on pikes or renouncing the current system. The former choice will come to bare as an inevitable reality should they ignore the problem long enough.

Another possibility to consider though is that jobs will simply not be automated despite the capacity to automate them. Corporations already keep a good 30% of the population in bullshit jobs that provide no productive value as feudal retainers in the US. The broader 50% of workers have some measure of unproductive bullshit in their workday even if their whole job is not bullshit. What's to say that our culture doesn't choose to make 90% of the population work bullshit jobs? How long can a culture maintain homeostasis when the majority of its population is operating in a stress based economy not because of need but purely out of hubris?

[firestar464]

I was able to reproduce Denis Xiloj's AGI graph with current data.



Damn missed Pi Day by one day LOL.

[caltrek]

An AI System Has Reached Human Level on a Test for ‘General Intelligence’
by Michael Timothy Bennett and Elija Perrier
December 24, 2024

Introduction:

(The Conversation) A new artificial intelligence (AI) model has just achieved human-level results on a test designed to measure “general intelligence”.

On December 20, OpenAI’s o3 system scored 85% on the ARC-AGI benchmark, well above the previous AI best score of 55% and on par with the average human score. It also scored well on a very difficult mathematics test.

Creating artificial general intelligence, or AGI, is the stated goal of all the major AI research labs. At first glance, OpenAI appears to have at least made a significant step towards this goal.

While scepticism remains, many AI researchers and developers feel something just changed. For many, the prospect of AGI now seems more real, urgent and closer than anticipated. Are they right?

Generalisation and intelligence

To understand what the o3 result means, you need to understand what the ARC-AGI test is all about. In technical terms, it’s a test of an AI system’s “sample efficiency” in adapting to something new – how many examples of a novel situation the system needs to see to figure out how it works.

Read more here: https://theconversation.com/an-ai-syst ... s-246529

[spryfusion]

Sam Altman's blog: "We are now confident we know how to build AGI"

We are now confident we know how to build AGI as we have traditionally understood it. We believe that, in 2025, we may see the first AI agents “join the workforce” and materially change the output of companies. We continue to believe that iteratively putting great tools in the hands of people leads to great, broadly-distributed outcomes.

We are beginning to turn our aim beyond that, to superintelligence in the true sense of the word. We love our current products, but we are here for the glorious future. With superintelligence, we can do anything else. Superintelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own, and in turn massively increase abundance and prosperity.

This sounds like science fiction right now, and somewhat crazy to even talk about it. That’s alright—we’ve been there before and we’re OK with being there again. We’re pretty confident that in the next few years, everyone will see what we see, and that the need to act with great care, while still maximizing broad benefit and empowerment, is so important.

[wjfox]