it literally got the first answer wrong. They still can't do arithmetic
although it's A LOT more impressive than others i seen
Last edited:
Artificial Intelligence thread
- Thread starter 9dashline
- Start date
it literally got the first answer wrong. They still can't do arithmetic
although it's A LOT more impressive than others i seen
have you used o1-preview for work on solving real work place automation issues? These things are far from achieving AGI.
The current line of AI development is all based on data and probability.
In certain aspects, it may appear in some use cases to be AGI, but given that it takes minutes to figure out how many r's are in strawberry, that should tell you where we actually are.
Thats cause cot reasoning models arent multimodal, attach a small vlm to it and it can see and count faster than you can
there is nothing inherently special about the biological brain....
R1 is slightly better... but weights not dropped yet, if its 16b instead of 32b, then game over for ClosedAI unless they already have AGI up their sleeves
Some people measure a lightbulb by energy consumed, some measure it by light output, some measure by light output divided by energy consumed.
Stanford ranked US #1 in money spent
It doesnt really matter if AGI isnt coming anytime soons, the amount of money US has spent on brute forcing this stuff, at the expense of every other research, means China now has a vector into wrecking the entire US tech and financial ecosystem by leveraging China's superior talent base to develop superior open source products that uses a fraction of resources and cost.
If OpenAI lowers prices they'll either go bankrupt or raise more money and suck even more life out of US tech industry. If they dont lower prices and convince companies to pay more, companies who uses it will themselves be financially disadvantaged compared to those who dont.
AGI can never be brute forced, the entire US approach is wrong, so the more money they throw into it the more f*ed they get
fundamentally, these models are based on probability. And there is only so far you can get with that. you won't even believe the amount of hallucination I get running o1-preview. I'd run something very obvious 10 times and it would give me the wrong answer 3 times.
Smallest unit of AI is the transformer... there is a paper about making the artificial nueron more complex instead of just scaling up network size. More scaling law directions and options are needed now...
Intelligence is substrate independent. AI chips and biological brains both have to follow laws of physics. Fundamentally the biological nueron is what then? Its all about the emergent property of complexity....its the emergence that is the intelligence, not the ReLU activation function or the fact that statistics is involved
China has two levers to pull... the rare earth and taiwan tsmc card if US AI gets too far ahead....
And the second lever is the dollar lever... China can out Fed the Fed now....the brrrrrrrr Printer finna run out of ink soon...
And all these free open weight models coming out will put US hopes of being entire worlds global AI as a Service to rest... Aint no one in 3rd world global south gonna pay ClosedAI a 30000% premium.... once China catches up in EUV, its over...
No more free lunch for USA
And if China gets to AGI first, well........
Application of AI is much slower than people think. I never quite understood what people think would happen if someone reached "AGI".
It's certainly great that we have so many open source options that people can use.
But at the present time, think of AI as a tool to improve productivity and automation. That's all it is.