Success is seductive, and the new hammer drives certain nails.
We’d do well to remember our #AI history, and that an algorithm is not a philosophy (or vice versa).
for those who think “deep learning is a philosophy” rather than a particular technique, what is that philosophy and is it, as the philosopher Popper would ask, falsifiable? what could as evidence against?https://twitter.com/dougblank/status/952558900699717632 …
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
1/2 The falsifiable insight of Parallel Distributed Processing (the 80s term) is that intelligence is an emergent property and could emerge from the collective behavior of simple units that are sparsely connected.
-
2/2 The Deep Learning falsifiable insight is adding some more structure to the units by clustering them in layers that are connected mostly sequentially will yield good results. Note in both cases the connection and properties of the units are not hand tuned to the problem.
End of conversation
New conversation -
-
-
I only really care about the argument of Intuition Machines vs Algebraic Minds. It isn't about all the arguments that you regurgitate again. It is about something fundamental that you continue to side step.
-
In addition, you fail to address the reality that there are many tribes of AI https://medium.com/intuitionmachine/the-many-tribes-problem-of-artificial-intelligence-ai-1300faba5b60 … and continue to promote the idea that its just one massive toolbox of algorithms. In short, you have no opinion about how AGI comes about.
- 1 more reply
New conversation -
-
-
Doesn't falsifiability per se apply to scientific theories and not philosophies? And deep learning is methodology/technique to arrive at an approximation to a function. Not sure it is any more falsifiable than a spline is as a function approximator.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Thomas Kuhn would say that deep learning is a paradigm, and paradigms aren't falsifiable. But if a new candidate paradigm can solve the problems that DL fails to solve, it will eventually replace DL as the dominant paradigm.
-
does it really classify as a paradigm tough?
End of conversation
New conversation -
-
This Tweet is unavailable.
-
-
The philosophy is that computation and decision making can be done by series of composed transformations in continuos space and both the representation into such a space and the transformations (useful of single or shared tasks) can be learnt; symb computing is its special case
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.