Dear Mohamed el Nawawy , interesting question! However, it seems to me that at many places you're attributing human traits to machines, and that these human traits for the machines could be achieved by what you claim is not needed, that is machine learning and optimization and learning by data. I think currently the approach is that knowledge acquisition in the form of data is not a problem (very large datasets are available), but explainability of the reasoning behind a certain conclusion (behind a certain selection of which data to use, if you like), is the problem. There's a new domain in AI nowadays called XAI, Explainable AI, that's looking into exactly this.

I hope the above is useful,

Best Wishes,

Haris

Dear Haris Shekeris,

Thank you for your contribution to this discussion. While I agree with you about the importance of AI explainability, I would like to clarify that my suggestion mainly involves a hypothetical case where the machine works through following simple instructions, eliminating speculations about machine’s reasoning capabilities upon which the concept of explainability is based.

The below link includes my thoughts about this:

https://medium.com/@melnawawy1980/speculations-about-the-end-of-ai-hype-a4cf63e3f085

Thanks,

Mohamed

Dear Mohamed el Nawawy ,

Many thanks for your answer and link to the post! I still feel a bit uncomfortable with the way you describe what AI is and why there is a hype. My discomfort I guess partly arises by your assertion that at the moment AI is trained with limited resources - i'm not sure that's the case. Either way, I'm also not sure that simple rules and many data would lead to good inferences - do you want to provide some examples of this? Do you think that the rules of inference and of thinking are actually very simple?

Best Wishes,

Haris

Dear Haris Shekeris ,

I’d mention the 1944 Colossus computer as an example of a machine which was programmed through plugboards and switches, which helped decipher the Lorenz encrypted messages during World War II.

Thanks,

Mohamed

Dear Mohamed el Nawawy ,

fair point, but then how much else could the Colossus do and how much can a modern smartphone do?

Best Wishes,

Haris

Dear Haris Shekeris ,

Thank you for this interesting discussion.

I agree that modern smartphones support a wide range of applications compared to Colossus which was a specialized computer. However, an essential trait distinctive to a machine in AGI is its ability to generalize its skills acquired in one domain to another which, for example, is not the case for AI chess programs due to their inability to apply their learned chess strategies outside the game.

Thanks,

Mohamed

Dear Mohamed el Nawawy , you're welcome,

Exactly, that's where may be the simple rules and plenty of data approach may fail, in the generalization aspect.

Best Wishes,

Haris