Ali Al-Ebrahim
@mi3law.bsky.social
Building with conviction in AI outside of deep learning since 2014. https://www.aolabs.ai/
I've fell in love with a new word-- broligarchs.
Like a shitty butterfly, techbros have morphed into broligarchs, and at long last I now have a word for it.
Credit to: x.com/jimstewartso...
Like a shitty butterfly, techbros have morphed into broligarchs, and at long last I now have a word for it.
Credit to: x.com/jimstewartso...
x.com
x.com
December 14, 2024 at 12:50 PM
I've fell in love with a new word-- broligarchs.
Like a shitty butterfly, techbros have morphed into broligarchs, and at long last I now have a word for it.
Credit to: x.com/jimstewartso...
Like a shitty butterfly, techbros have morphed into broligarchs, and at long last I now have a word for it.
Credit to: x.com/jimstewartso...
How is transformer architecture sufficient for AGI given that a model's weights are fixed?
Are our neural weights fixed as human GIs?
Deep learning happens through adjusting weights, which are then fixed during inference. But humans do learn, right?
#DeepQuestions
Are our neural weights fixed as human GIs?
Deep learning happens through adjusting weights, which are then fixed during inference. But humans do learn, right?
#DeepQuestions
November 27, 2024 at 4:17 AM
How is transformer architecture sufficient for AGI given that a model's weights are fixed?
Are our neural weights fixed as human GIs?
Deep learning happens through adjusting weights, which are then fixed during inference. But humans do learn, right?
#DeepQuestions
Are our neural weights fixed as human GIs?
Deep learning happens through adjusting weights, which are then fixed during inference. But humans do learn, right?
#DeepQuestions
Large AI models are (pre) trained on inhuman amounts of data yet still fail common sense.
Humans need far less training because we can learn.
What's next in AI seems simple when put that way, provided one can look beyond the angels of hype and scale.
Humans need far less training because we can learn.
What's next in AI seems simple when put that way, provided one can look beyond the angels of hype and scale.
November 25, 2024 at 8:21 AM
Large AI models are (pre) trained on inhuman amounts of data yet still fail common sense.
Humans need far less training because we can learn.
What's next in AI seems simple when put that way, provided one can look beyond the angels of hype and scale.
Humans need far less training because we can learn.
What's next in AI seems simple when put that way, provided one can look beyond the angels of hype and scale.