Yes big data is good, Google sorta 'is' big data. The Hutter Prize / AGI 
research is best focused on a small dataset to compress data and really doesn't 
matter if its vision or text. Scaling is done later. Google is not all that 
focused on AGI 100% then, as it serves us and stores lots of data, of course it 
has a big team though and I believe they are working hard on something like AGI 
. . . . just they don't seem to have the brains (billions of dollars later) to 
*NOTICE *the Hutter Prize and show a lossless compression score using 
Transformers. I don't care why Transformers can't compete, I know they can! 
I've went through it in my head 10 times "ok Transformers make a huge model, 
etc, but really they can compete at the Hutter Prize". In the HP you typically 
start with the code and it grows the model as it compresses, then stores the 
'error correction code' and trashes the model, then upon decompressing it back 
perfectly to original dataset it again grows the model as it pops out each 
bit/byte and finally trashes the model again along with the 'error correction 
code' and the AI code, if you will. The idea is the error correction code 
stores it compressed, the model at each step makes up for the missing data 
because it predicts what the data is pretty correctly.

Why again can't Transformers compete? And I mean why 'impossible', not why they 
just don't i.e. they "could" use CPU, wait for it to run, etc.

I disagree Matt, AGI is the way far forward. No matter how many humans work 
together, which, we can't seemingly, we still would not think 100 times faster, 
be more intelligent, clone adult AGI brains and differentiate them to do 
similar tasks in paralell, etc. The intelligence upgrade is a big one, you 
can't even add 9 trillions humans or data or speed to Earth to match it, it is 
just a leap ahead of them, proven by looknig at tests on the Hutter Prize. It's 
the first thing you see in tests, adding more data makes your compressor 
compress 100MB to ex. 30MB, then more data 28MB, more data 27MB, more data 
26.8MB, more data 26.7999999.....then you drop the intelligence upgrade and get 
to 21MB! Clearly intelligence is gazillions of times more important than any 
other advantage artificial intelligence has such as speed in thinking or 
motors, data size / brain size, cloning adult AGI brains, never dying, new 
sensory cortexes, higher resolution sensors, more digital redundancy and same 
for motor precision, etc.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Tee75800a5ab37f5b-Mfbbdc3b148a4e167e2d4eba1
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to