Re: Even Google AI Overview understands me now --- different execution traces have different behavior !!!

On 2024-10-03 02:09:39 +0000, olcott said:

On 10/2/2024 5:48 PM, Richard Damon wrote:

On 10/2/24 10:39 AM, olcott wrote:

On 10/2/2024 6:08 AM, Richard Damon wrote:

On 10/1/24 7:26 PM, olcott wrote:

On 10/1/2024 12:58 PM, joes wrote:

Am Tue, 01 Oct 2024 12:31:41 -0500 schrieb olcott:

On 10/1/2024 8:09 AM, joes wrote:

Am Tue, 01 Oct 2024 07:39:18 -0500 schrieb olcott:

On 10/1/2024 7:19 AM, olcott wrote:

https://www.google.com/search?
 

 

q=Termination+Analyzer+H+is+Not+Fooled+by+Pathological+Input+D&sca_esv=889093c5cb21af9e&sca_upv=1&source=hp&ei=Muf7ZpOyMZHfwN4PwYL2gAc&iflsig=AL9hbdgAAAAAZvv1Qg04jNg2ze170z3a8BSGu8pA29Fj&ved=0ahUKEwiTk7zkk-2IAxWRL9AFHUGBHXAQ4dUDCBg&uact=5&oq=Termination+Analyzer+H+is+Not+Fooled+by+Pathological+Input+D&gs_lp=Egdnd3Mtd2l6IjxUZXJtaW5hdGlvbiBBbmFseXplciBIIGlzIE5vdCBGb29sZWQgYnkgUGF0aG9sb2dpY2FsIElucHV0IERIAFAAWABwAHgAkAEAmAEAoAEAqgEAuAEDyAEA- 

AEC-AEBmAIAoAIAmAMAkgcAoAcA&sclient=gws-wiz

https://chatgpt.com/share/66fbec5c-7b10-8011-9ce6-3c26424cb21c

It sounds like it’s trained on your spam. LLMs don’t know anything
anyway.

 

I asked: „What happens when HHH tries to simulate itself?”
ChatGPT: [my comments in brackets]

 

### 1. **Self-Invocation in HHH**
If `HHH` calls itself (i.e., `HHH(HHH)`), it creates an infinite loop
of calls unless there is some form of termination condition built into
`HHH`.
Without such a condition, `HHH` would never return, resulting in non-
termination.
[So not a decider.]
 ### 2. **Halting Criteria**
In a well-designed termination analyzer, `HHH` would need to have
specific logic to prevent endless recursion:

 

[Not present.]
 ### 3. **Practical Consequences**
If `HHH` lacks the ability to handle self-referential calls correctly:
- **Stack Overflow**: Continuous self-invocation will eventually
exhaust the stack memory, leading to a stack overflow error.
[Not considered in theory.]
- **Incorrect Results**: If not designed carefully, it may lead to an
incorrect conclusion about halting, misrepresenting itself as either
halting or non-halting.
 ### 4. **Best Practices for Termination Analyzers**
To avoid these pitfalls, a termination analyzer should implement:
- **Guard Clauses**: Checks at the beginning of the function to prevent
unnecessary self-simulation.
- **State Management**: A way to track the functions being analyzed,
allowing `HHH` to recognize if it's being invoked with itself and act
accordingly.
 ### Conclusion In summary, if `HHH` tries to simulate itself without
proper safeguards,
it risks falling into an infinite recursion or causing a stack
overflow. A robust design would include checks to prevent
self-invocation, ensuring the termination analyzer can handle all
cases, including its own function,
gracefully.
 

*You can continue this conversation with ChatGPT*

I’ll leave that to you.

Click on the link and see how it answers this question:
Is H a Halt Decider for D?

You should feed it our objections.
If you believe in it only when you prompt it, it is not suited as an
authority (fallacious anyway).
 

 You feed it your objections.
It will tell you how and why you are wrong.
 

 No, it will tell you something that matches the words you told it.
 You don't seem to understand what Large Language Models are.
  You seem to forget that LLM know nothing of the "truth", only what matches their training data.
 They are know to be liars, just like you.

 In the case of their evaluation of my work they are correct.
 *Try this yourself*
 https://chatgpt.com/share/66fbec5c-7b10-8011-9ce6-3c26424cb21c
 Does HHH have to abort its emulation of DDD to prevent the infinite execution of DDD?
 

 Try asking it
 Why is it ok for HHH to say that DDD doesn't halt when it will when run?
 

 You guys try to pretend that the pathological relationship between
HHH and DDD does not make any difference to the behavior of DDD
knowing full well that it does make a difference.

The behaviour of DDD is what DDD does if executed. As DDD takes no input
its behavour is always the same. What does "make a difference" mean
in this context?
--
Mikko