Is an AI really a Turing Machine ?

[xkalibur08]

Great question!

Well, while an AI does have some similarities to a Turing Machine, it's like comparing apples to oranges, ya know? A Turing Machine is a theoretical concept, while AI is all about real-world applications and learning from data. So, they're similar in some ways, but AI takes things to a whole new level with its ability to adapt and improve. 

Hope that clears things up for ya!

Cheers!

[Genady]

34 minutes ago, xkalibur08 said:

Great question!

Well, while an AI does have some similarities to a Turing Machine, it's like comparing apples to oranges, ya know? A Turing Machine is a theoretical concept, while AI is all about real-world applications and learning from data. So, they're similar in some ways, but AI takes things to a whole new level with its ability to adapt and improve. 

Hope that clears things up for ya!

Cheers!

What makes you think that Turing Machine cannot be applied to real world, cannot learn from data, adapt and improve?

[Ghideon]

On 5/31/2023 at 3:49 PM, studiot said:

I know the official line is that so called AI programs are nothing more than glorified TMs

But is that really true ?

 

My intuitive answer would be that CHATGPT and similar could be implemented on a Turing Machine. But here are some thoughts; hopefully others may add some insights.

Assumptions that as far as I know are sound in this context:
1: CHATGPT and similar systems are build from components that can be realised by Turing machines.
2: The architecture of CHATGPT (and others) is based on distributed components and interacting concurrent computational resources.
3: Frequent changes occur through intentional updates and upgrades, unintentional failures, user interaction etc.

There are other computational models than Turing, for instance Hewitt's Actor model of computation.

Short description*

Quote

The model has been used both as a framework for a theoretical understanding of concurrency, and as the theoretical basis for several practical implementations of concurrent systems. The advent of massive concurrency through client-cloud computing and many-core computer architectures has galvanized interest in the Actor Model.

Question: Is the Actor model of computation more suitable than Turing machine for a large distributed system such as CHATGPT? 
I don't not know; assume for this discussion that the answer is "yes" and move on to one of Hewitt's claims*

Quote

There are nondeterministic computable functions on integers that cannot be implemented by a nondeterministic Turing machine.

This could mean that there may be more to @studiot's initial question than I initially thought (hence my late answer in this thread).
Your thoughts? 

 

*) https://arxiv.org/pdf/1008.1459.pdf. 
(I chose to not go into details about the differences between the models and their applicability in this post)

 

Edited June 29, 2023 by Ghideon
grammar

[studiot]

5 hours ago, Ghideon said:

My intuitive answer would be that CHATGPT and similar could be implemented on a Turing Machine. But here are some thoughts; hopefully others may add some insights.

Assumptions that as far as I know are sound in this context:
1: CHATGPT and similar systems are build from components that can be realised by Turing machines.
2: The architecture of CHATGPT (and others) is based on distributed components and interacting concurrent computational resources.
3: Frequent changes occur through intentional updates and upgrades, unintentional failures, user interaction etc.

There are other computational models than Turing, for instance Hewitt's Actor model of computation.

Short description*

Question: Is the Actor model of computation more suitable than Turing machine for a large distributed system such as CHATGPT? 
I don't not know; assume for this discussion that the answer is "yes" and move on to one of Hewitt's claims*

This could mean that there may be more to @studiot's initial question than I initially thought (hence my late answer in this thread).
Your thoughts? 

 

*) https://arxiv.org/pdf/1008.1459.pdf. 
(I chose to not go into details about the differences between the models and their applicability in this post)

 

I don't think anyone has a full handle on AI  so thank you to all members who have added their piece.

+1 to Ghideon for a particularly cogent offering.

 

I have certainly learned many things from this thread, including how many 'Turing Machines' or variations thereof there are.

I was only thinking of the most basic one when I started this.

 

I do feel there is plenty more  good mileage left in this subject yet.