(23-11-2025, 02:38 PM)Digitalgoldfish79 Wrote: You are not allowed to view links. Register or Login to view.

(23-11-2025, 02:31 PM)Philipp Harland Wrote: You are not allowed to view links. Register or Login to view.Research is the process of identifying a problem and then using relevant, applicable methods to resolve it. AI cannot write and debug code, especially not for this purpose. The graphs are a result of hallucinating. 

Let's try an example with something in my field. As I am a mathematician by trade, that field is mathematics, obviously.

Let's say we have a hypothetical family of differential equations, D. Let's say D is vastly uncharted in terms of mapping out its properties and relations to other families of differential equations, i.e. there have only been 1-2 papers on D in the past 10 years since its discovery and those papers haven't really made much progress.

AI couldn't be of any use, especially not for writing and debugging code that could help us understand D.  

That's the fatal flaw. You've probably heard this before, but AI cannot produce novel information. It's just a numbers-in numbers-out machine.

I understand AI very well, and it's perfectly capable of suggesting appropriate statistical techniques, and then writing code to help you do execute them. The work I did was purely in Termux using python. We start with EVA and go from there. What's the difference between that and any other computational statistics?  What's not right is where people try to operate entirely within the Chatgpt environment. That's fatal as it will hallucinate, forget, and generally create a mess. I have python scripts and tsvs for every single calculation in my paper, but I admit my figures are broken which is highly embarrassing and I will fix them

(23-11-2025, 02:49 PM)Digitalgoldfish79 Wrote: You are not allowed to view links. Register or Login to view.

(23-11-2025, 02:47 PM)Philipp Harland Wrote: You are not allowed to view links. Register or Login to view.

(23-11-2025, 02:38 PM)Digitalgoldfish79 Wrote: You are not allowed to view links. Register or Login to view.

(23-11-2025, 02:31 PM)Philipp Harland Wrote: You are not allowed to view links. Register or Login to view.Research is the process of identifying a problem and then using relevant, applicable methods to resolve it. AI cannot write and debug code, especially not for this purpose. The graphs are a result of hallucinating. 

Let's try an example with something in my field. As I am a mathematician by trade, that field is mathematics, obviously.

Let's say we have a hypothetical family of differential equations, D. Let's say D is vastly uncharted in terms of mapping out its properties and relations to other families of differential equations, i.e. there have only been 1-2 papers on D in the past 10 years since its discovery and those papers haven't really made much progress.

AI couldn't be of any use, especially not for writing and debugging code that could help us understand D.  

That's the fatal flaw. You've probably heard this before, but AI cannot produce novel information. It's just a numbers-in numbers-out machine.

I understand AI very well, and it's perfectly capable of suggesting appropriate statistical techniques, and then writing code to help you do execute them. The work I did was purely in Termux using python. We start with EVA and go from there. What's the difference between that and any other computational statistics?  What's not right is where people try to operate entirely within the Chatgpt environment. That's fatal as it will hallucinate, forget, and generally create a mess. I have python scripts and tsvs for every single calculation in my paper, but I admit my figures are broken which is highly embarrassing and I will fix them

I'm curious now. I'd like to test your knowledge about AI. Explain how an LLM works without  using Google or LLMs themselves. A summary in 200 words or less would be nice.

Don't be an arse

(23-11-2025, 02:23 PM)Digitalgoldfish79 Wrote: You are not allowed to view links. Register or Login to view.Thank you.. I stupidly uploaded the wrong diagrams... I'll recreate them in python

(23-11-2025, 02:52 PM)Ruby Novacna Wrote: You are not allowed to view links. Register or Login to view.

(23-11-2025, 02:23 PM)Digitalgoldfish79 Wrote: You are not allowed to view links. Register or Login to view.Thank you.. I stupidly uploaded the wrong diagrams... I'll recreate them in python

Oh really? Because there are right and wrong diagrams in your 36-page article?

P.S. Why do you keep repeating the word Python?
Is it some kind of magic word that solves all problems?

(23-11-2025, 02:58 PM)Digitalgoldfish79 Wrote: You are not allowed to view links. Register or Login to view.I hadn't realised how paranoid everyone was about AI... I'm head of digital and AI for a bank and we have teams of people using it every day to write code and do statistics.. this is all pretty normal to me...

(23-11-2025, 03:11 PM)dashstofsk Wrote: You are not allowed to view links. Register or Login to view.Instead of just stating that Chi-squared and Fisher tests on the data showed a lack of independence and randomness it would have been useful to present some definite examples from the manuscript. In order for us to judge whether you have applied these methods correctly you really do need to give us more of your calculations.

My own efforts to apply statistical hypothesis testing techniques on the text of the manuscript has raised many anomalies. A lot of data is many standard deviations away from what would be expected. It just highlights what we know already, that the text is irregular and doesn't have an easy explanation.

(23-11-2025, 03:12 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.[quote="Digitalgoldfish79" pid='74463' dateline='1763906309']
I hadn't realised how paranoid everyone was about AI... I'm head of digital and AI for a bank and we have teams of people using it every day to write code and do statistics.. this is all pretty normal to me...