This is why AI may never surpass human brain intelligence.

This treats intelligence as a hardware spec race (ops per second, watts, plasticity) and then concludes that because the brain wins some of those races today, AI can’t reach human-level intelligence. Intelligence isn’t a FLOPS competition. The question of whether AI achieves general intelligence is entirely decoupled from whether a GPU currently beats a neuron on power efficiency. These are different levels of description entirely. It’s like arguing a microphone can’t hear because it has no eardrums, or a camera can’t see because it has no retina. And “never crashes” is false on its face. Strokes, seizures, psychosis, sleep deprivation, cognitive bias, confabulation, motivated reasoning. The brain “crashes” constantly just in ways we normalize or don’t call “crashes” because they’re familiar. Using current-state hardware comparisons to make claims about fundamental limits is juvenile and brutally simplistic. A proper assessment of intelligence looks at the results (has conversations, sees what it predicts), it doesn’t pretend intelligence is defined by specific pieces doing specific things. That’s pure fallacy.

📢Buy The Book Now👇 The Borrowed Mind by @JohnNosta 💡The era of #AI #LLMs 👉Reclaim Human Thought 🧠

We are a biological miracle and yet here we are destroying ourselves 🥲

Ai can only learn to be like us they can never be more than us. Because they don’t have the air we breath

total bio privilege nonsense there's zero principled barrier in physics or engineering to silicon (or photonic, memristor, or whatever-next-gen) hardware reconfiguring connections orders of magnitude faster. History is littered with never pronouncements that tech proved wrong

It will and easily, simply because AI has instant recall, and humans don't. Our only hope is symbiosis.

Five Nvidia H100 cards provide 1 quadrillion operations and can fit into a large PC case equipped with a workstation motherboard. The total power consumption for the PCIe cards is 1,750 watts (5 × 350W) plus another 400-600 watts, for a range of 2.3kW to 2.5kW under full load.

@grok, estimate the year when AI will reach superhuman intelligence at current rates. Also list some bottlenecks to solve in order to accelerate towards this goal.

lol video is tarded Just xai's is 1000-10000x human level inteligences, sure the power efficiency needs 1000x work but it'll get there, and there's multiple areas of rapid improvement. RT viralrush you get bad info.

Current AI is a crude imitation of part of our brain function. Its an energy inefficient force hack. Yet its quite impressive and it will get exponentially better. Yet a long way to go come to the sophistication and complexity of the human brain which we only understand superficially.

🙏

This is the point, where most computer scientists fail.

AI will be going to space along with the data centers. That's when it will excess humams

Apparently it needs pointing out that “never” encompasses billions upon billions of years into the future. Throughout the infinite complexity of this universe. Even, perhaps, universes beyond this one. We don’t know what will happen tomorrow. Have some humility.

Lord Kelvin : "Heavier-than-air flying machines are impossible". George W. Melville : "The pursuit of manned flight is a vain fantasy" Simon Newcomb : "Flight by machines heavier than air is unpractical and insignificant, if not utterly impossible". Same mentality...

Never crashes in a bit optimistic.

Exactly and that massive energy consumption for data centers is seriously ridiculous

Ai is what was said to be one of the things that will happen during the end times and we are already living in it. Most people don’t see this

It might be true, but quite a logical fallacy - "because human brain is complex and efficient etcetc., AI is not" Like 10 years ago we didn't even had good RNNs and see the state of modern LLMs going from bad at basic physics to solving PhD level math problems.

Small omission: an unmyelinated neuron transmits information at roughly walking speed (on the order of m/s). Modern computers run at GHz (10⁹ ops/s) and signals propagate near the speed of light (~10⁸ m/s). The biggest gap, though, is context of awareness: humans can hold ~4 (albeit complex) items in conscious context; LLMs now operate with million items, and that number is growing rapidly.

This is comforting 😊

Its not being. It is a machine and is synthelligent - no subjective experience. It is only of value if it empowers being or life. Otherwise it's useless - redundant noise. Of course if we direct it to only process data that empowers being, it becomes a brilliant tool.