Compute vs Thinking
Machine Compute or Thinking?
The question in the subtitle is quite provoking, right? It is a question I have been pondering for some time now if I want to figure out other questions about machines, such as consciousness, understanding, etc. So…does the machine we built compute or think?
Based on Webster's Dictionary, here is the meaning between the two.
Think: to exercise the powers of judgment, conception, or inference, form opinion, reflection
Compute: to determine especially by mathematical means
Now if you are to strip down AI in its past and current form, it is operating on “compute” rather as we humans are the one that sets down the rules, algorithm and models that are maths or logic based rather. You can explicitly write down the computation model if we want to. Why the confusion and perplexity is because we are using deep learning model or neural network models. The usage of neural network models has brought about the advancement of AI but it has also introduced a lot of uncertainty on how we understand machines, together with the human bug call “anthropomorphism”. These two potent factors formed a mixture that made experts even more confused on AI, from what I see.
I hope to keep the newsletter open to as many as possible. :) If you have benefited much from my posts and will like to join my cause, consider making a “book" donation. Link at the bottom of the newsletter or here. :)
Simply put, neural network is the black box we put into machines, to give them “intelligence” but we humans got so confused as we do not have the minute detail on how the black box works and thus felt that AI is dangerous, a normal behaviour in how humans deal with uncertainty. AI will not act outside of the black box because they compute, they do not think out of the box, literally speaking!
This is why I feel that it is time for us to move out of the deep learning architecture and start thinking about a connectionist core but a symbolic model layer on top of it for better regulation on the decision the AI will make, going back to the previous newsletter I have written on.
What are your thoughts? Please share them in the comment or Chat! :)
Recommended Content
What is the one thing that is often overlooked in the overlooked Data Management? <Post>
Want to try webscraping to collect data but do not know how to start? <Podcast>
Want to know more about MLOps? Thu Ya and I discuss it together with Ryzal Kamis <Podcast>