AGI might be soon, felt very suicidal.

[Archness]

I hate AI, I absolutely hate it, axiomatically. It's "helpful", but in the end humans will only use it to extract exploit and replace each other. This thing to supersede the place of humans on this earth, how they're make our true insignificance within the universe undeniable, un-ignoreable, and absolutely unacceptable... I hate it so, I'd destroy all computers and the internet to prevent or atleast postpone the coming of advance AI. Even if I'd die in the process, as addicted to all this junk I am, I'd do it. I'm tired of everyone doing all this gaslighting about how nothing will magically happen and that AI will be good. NO, at best I'd reduce humans to cattle eating grass all day as the farmer "machine god" takes care of everything, or more realistically 98% of people are left behind in destitution with no real power anymore as the "1%" take over everything.

I hate it, I hate it, I hate it. But I'm powerless as it's developed, as everyone just accepts it as "new tech" and adopt this learned helplessness for whatever new tech happened to get cooked up. There's only blind faith and waiting. There's nothing I can do, I won't even be able to find a job or do anything because AI would be doing everything and would just take up any place a human could be in. I've herd an insider talk about how close we are getting to AGI and I hate it. Every time I hear such news, I feel how this whole world is moving, how stuck I am with to to just endure suffering and this nihilism only for the light at the end to be death itself.

All of this really makes me suicidal.

 

[destinationlosangel]

100%. I myself have played around with many of these AI LLMs and it is scary AF. I usually am optimistic about tech but not in this case. AI is gonna change everything for the worse

 

[Darkover]

artificial general intelligence is decades away 30-50+ years away

The adult human brain weighs about 3 pounds (1,300-1,400 g).
The adult human brain is about 2% of the total body weight.
The average human brain is 140 mm wide. 140,000000
The average human brain is 167 mm long. 167,000000
The average human brain is 140 mm height. 140,000000

3,273,200,000,000,000,000,000,000 nm in the human brain

1 mm equals 1 million nanometers
1 meter equals 1 billion nanometers

Apple M2 Ultra SoC: As of June 2023, this ARM-based dual-die SoC had the highest transistor count of any consumer microprocessor at 134 billion.

in the world's most powerful supercomputer

Taihu Light supercomputer in China that is currently ranked as fastest (based on the Linpack benchmark) has about 40,000 of those processors. So, about 400 trillion transistors in the processing part of the hardware.

But like most computers, the bulk of the transistor count is in the main memory, not the processors. Most of the 1.3 petabytes of memory in the Taihu Light is DRAM, and each bit of DRAM memory takes one transistor per bit. A byte takes 9 bits, not 8, if you include the error-correction bits. "Peta" means 10 to the 15th, or "a quadrillion" if you prefer an English name for the number. So the DRAM includes about 12 quadrillion transistors, and that's about 97 percent of all the transistors.

thats 12,000,000,000,000,000 transistors in its memory and 400,000,000,000,000 transistors

that's a total of 12,400,000,000,000,000 transistors in the world most advance super computer
vs human brain 3,273,200,000,000,000,000,000,000 nm in the human brain

that means the world most powerful super computer could fit inside of the space of a human brain 263,967,741 million times

 

[ssspadbye]

Darkover, you're taking an incredibly simplistic approach to AGI. The idea of AGI is certainly not for a single chip to replace the human brain, or even act exactly like one. AGI is certainly nowhere nearly 30-50 years ago. Is suffices to see the logarithmic curve of development - AGI will almost come out of nowhere (and soon) and in the blink of an eye the world will change forever.

Archness, I'm very, very informed on this topic and share your concerns. The world as we know it today, the value of humans in it, the systems that are the bedrocks of society - all of it will change in a blink of an eye, and in the not too distant future. There is no slowing or controlling this progress, because the world is no longer practically monopolar as it once was. If for example the US doesn't do it, China will. If not China, India. And also there is more and more being made accessible via open source, coupled with countless billion dollar enterprises will access to cheaper and cheaper mass cloud computing (with lightning-pace developments from the likes of Nvidia).

The fact is, we cannot stop this progress. Either we live to see humans become second-class citizens of a society of our creation, or we die before it. For me, I feel it may be the latter, although I'm not 100% certain.

 

[Darkover]

Darkover, you're taking an incredibly simplistic approach to AGI. The idea of AGI is certainly not for a single chip to replace the human brain, or even act exactly like one. AGI is certainly nowhere nearly 30-50 years ago. Is suffices to see the logarithmic curve of development - AGI will almost come out of nowhere (and soon) and in the blink of an eye the world will change forever.

The human brain and supercomputers like China's Taihu Light indeed differ vastly, both in size and processing approach. my comparison between the brain's nanometer-scale complexity and the transistor count in supercomputers offers an intriguing perspective.

For context, the human brain's neural structure and connectivity density are immense: each neuron has thousands of synaptic connections, leading to trillions of pathways for processing information. A supercomputer, on the other hand, has an incredible number of transistors for raw data processing, but it lacks the complex, interconnected architecture of neurons and synapses that the brain has.

Let's break down some of the figures:

1. Brain Size vs. Supercomputer Size:
   • With the brain weighing about 3 pounds and having approximately 3.2×10243.2 \times 10^{24}3.2×1024 nanometers in volume, it's small compared to the world's largest supercomputers, which take up entire rooms.
   • If we're discussing sheer volume, your calculation that the Taihu Light's transistors could fit inside the brain around 263 billion times demonstrates just how much more compact and energy-efficient the brain is.
2. Transistor vs. Synaptic Count:
   • The Taihu Light supercomputer has around 12.4 quadrillion ( 1.24×10161.24 \times 10^{16}1.24×1016 ) transistors.
   • The human brain has an estimated 86 billion neurons, each connecting to thousands of other neurons. This leads to around 1×10151 \times 10^{15}1×1015 synapses, which are complex, adaptable, and play roles in memory and learning through electrical and chemical signals — something even the most advanced transistors don't replicate.
3. Processing Power & Energy Efficiency:
   • Current supercomputers are optimized for different tasks (e.g., large-scale computations, simulations) but aren't structured for general intelligence or adaptable, flexible learning as the human brain is.
   • The brain operates on about 20 watts of power, while supercomputers require megawatts.
4. Implication for AGI Development:
   • AGI would likely require not just computational power but an architecture that can dynamically adapt, learn, and self-modify, similar to the brain's structure.
   • Given current technological constraints, this unique processing flexibility and efficiency may indeed place AGI decades away.

The supercomputers are optimized for their current tasks, while the brain's structure has evolved specifically for perception, adaptation, and memory, making a direct comparison difficult beyond these metrics. So, while the hardware in modern supercomputers like the Taihu Light has raw power, they're still a long way from matching the efficiency and adaptability of the human brain, suggesting that true AGI might require advances in hardware and architecture on top of raw processing power.

 

[ssspadbye]

I feel this is apples to oranges, comparing a singular human brain to a connected intelligence model based on a cloud infrastructure.

 

[Darkover]

I feel this is apples to oranges, comparing a singular human brain to a connected intelligence model based on a cloud infrastructure.

we don't even have fully automated self driving cars and there not expected to arrive until at least 2035
yet some claim AGI is just around the corner

 

[Alpenglow]

I'm not sure what you mean by AGI? The closest thing right now would be Q* from OpenAI which is more of a rumour and may be a math algorithm.

I agree we'll probably get better models in the coming decades for whatever tasks the large AI companies decide. The limiter so far seems to be data, processing power, and time. The respective error curves of which seem to have some flatten out. This is likely due to the current design being predictive and based on probabilities where there is unavoidably some error.

I imagine you'd need to make a model able to be logical and manipulate ideas, which would be difficult since error becomes significantly more difficult to quantify. Though if you did want to start somewhere, you would start with a math algorithm like Q* (if it exists) since math is logic basically.

I agree that AGI will probably come in the future, but we need multiple breakthroughs for that to happen. And even if countries are hiding their research, they are probably not more than one or two steps along the path to AGI. Although the contenders are probably just China and the US, and any research is probably discovered by their respective intelligence agencies. No other country has the tech companies necessary, and no (concealed) amount of government research is enough to outpace private tech.

 

[alienfreak]

I have academic qualifications in this field (not to imply that it means much) and would say there is no sign that we are anywhere near AGI. I think you guys are being swept away by propaganda, to be direct.

Neural networks can be used to make horrific killing machines and perform mass surveillance into all our thoughts, yes, but it is just tools for humans. There is no sign of a machine god with any reasoning capabilities, at this point.