AI Hype-o-Meter
A lot of problems here.
1. The human brain, according to many sources, is estimated to have up to 1000 trillion synapses.
2. The comparison between human neurons and artificial neurons is deeply flawed as biological ones are very complex and their functioning is poorly understood.
3. The relationship between the number and synapses and computing power is gratuitous since we don’t know much about how the brain works and so there is no way to relate what it does with any metric applied to computers.
The only way to compare dissimilar processors is to see if these create comparable results.
The Cerebras chip is aimed at supporting larger language models like GPT-3 which was trained using a mere 172 billion machine learning parameters.
Anyone who knows how these language processing apps work knows that they have absolutely no understanding of what the words they are generating mean.
They have no understanding of anything because they are mindless algorithms, kind of like an electronic parrot.
Humans, on the other hand, don’t generate text, they speak and read, they understand what the words mean.
No doubt the Cerebras chip can be used to create even bigger electronic parrots but making comparisons to the human brain is sheer hype.
