Is AGI really impossible I know even Even if someone suddenly discovered the algorithm for AGI, which I don't know how is possible/impossible.
There still would be many Physical Limitations.
If algorithm determines what kind of intelligence is possible physics will determine how much computation can actually be carried out.
>>161939
I might sound midwit, (which I actually am)
But don't they need to build the physical setup first, then try running algorithms and experimenting eith them
Every work done by human can be replaced by ai economically.
Let's go with this as a definition.
And yes my prediction for AGI was 2028
<Why?
Not Because we will have a good algorithm but becuz the compute will be astronomical.
For eg the compute which will come online will be so much that a 2024 6 months training run will take just 3 hours In raw compute term forget the adjustment and advancements in the training run.
And which the predicted date is wrong because the AGI is already coming in the December itself.
The next year is full of agi-era war.
2028 we many lose this world to robots.
>>161954
Hardware and algorithms evolve together. Current GPU/TPUs can run today's AI, but a more advanced algorithm might need completely different hardware, such as neuromorphic or optical computers. So even if we discover the algorithm, we may still need to build the physical infrastructure capable of running it efficiently.
Can anyone explain what kind of Algorithm we actually need and how it have to different from AIs one thi g I k ow it should have tendency to self learn, self adapt, self re-code itself
>>161939
idk yaar, maybe have different models break tasks down into smaller chunks and ultimately feed the output to a bio hybrid computer that makes the decisions
>>161997
There's a pure bio-compute available for rent in australia https://corticallabs.com/https://corticallabs.com/
If this is what you have in mind
