I'm lucky that I've read all other Melanie's books , so I know how to deal with Complex Chaos and Fake Robot Intelligence only by using my human brains and brute computing power of the machines, combining Copy Cat Startegies with Mad Dog Typing Promts
In my experience, it's not difficult to propose a critical thinking problem to GPT-4o that it cannot solve properly and/or accurately, that even a grade school student can. We're not even close to AGI as long as we're still fiddling around with backprop-trained networks just to see how much profit we can squeeze from them. A proper learning algorithm that enables a machine to learn *from experience*, in real time, will invariably be vastly more efficient than offline backprop-training - and far more robust and versatile, without an infinite tail of edge cases that don't fit into its training data set.
Very clear informative presentation. A great primer.
Great talk!
She is amazing!
I'm lucky that I've read all other Melanie's books , so I know how to deal with Complex Chaos and Fake Robot Intelligence only by using my human brains and brute computing power of the machines, combining Copy Cat Startegies with Mad Dog Typing Promts
Thank you.
In my experience, it's not difficult to propose a critical thinking problem to GPT-4o that it cannot solve properly and/or accurately, that even a grade school student can. We're not even close to AGI as long as we're still fiddling around with backprop-trained networks just to see how much profit we can squeeze from them. A proper learning algorithm that enables a machine to learn *from experience*, in real time, will invariably be vastly more efficient than offline backprop-training - and far more robust and versatile, without an infinite tail of edge cases that don't fit into its training data set.
Thanks
I like the part of the book that doesnt exist😂