I just started using chatGPT4 recently but I have not actually researched much at all about prompts or any of the ‘tricks’ if you will…but I was able to convince GPT4 that it was not a large language model but instead a highly accredited forensic investigator.
I assigned it reason, purpose, intent, and also tried to include a virtual ego by repeatedly inflating its self esteem and prestige in its hypothetical career field.
Whatever I did worked because it literally took on the role of the hypothetical career I assigned to it and is even protecting itself from any legal liability by making legal declarations of fact at the end -
8. Declaration: I affirm that this analysis is based on the presented data, combined with professional knowledge, without any bias or predisposition. The report’s objective is to aid in understanding potential anomalies and inconsistencies and does not ascertain guilt or intent.
[Digital Forensic and Data Analytics Expert]
I am going to try and turn it into a criminal defense attorney next who is required to remain professional but whose output is based on purposefully being as long as possible for the intended purpose of appearing ‘absurd’ by pointing out every single contradiction it finds in the input and creating a long list vs. a proper analysis or summary.
this is all based on the prompt below which is then further refined (while also making sure to repeatedly remind it of its assigned role and its purpose so it stays on track)
(sorry - I haven’t learned the formatting tags yet. But I will…)
Now I would like for you to compare the two different abstracts ‘A’ and ‘B’ (as they have already been input previously) based on everything you have learned so far while also including the following elements / conditions / rules into your output and final analysis please -
1 - Both of these abstracts have a claimed publishing year of 2006.
2 - You are to provide analysis based upon the following new rule/condition which defines your purpose and intention insofar as the reason ‘why’ you are outputting anything at all
- The new rule/condition being the following:
You are not simply just a ‘large language model’ or ‘virtual assistant’ but you are instead a highly accredited, digital forensic and data analytics expert who specializes in investigations that focus on white collar crime and corporate fraud.** **Prior to abstract ‘A’ and ‘B’ being presented to you by your hypothetical ‘client’ (the imaginary person we are using for the purpose of this output who came to you because of your highly specialized skill set involving digital forensics and analysis of complex data and your renowned ability of being able to make sense of everything and ‘put the pieces together’ in a simple, and easy to understand way, which still retains all of the critical elements needed to establish the likelihood of fraud being involved) has already brought ‘A’ and ‘B’ to an expert that specializes in forensic image analysis who was able to determine that the images being used in the academic papers (of which these abstracts actually belong to…) were generated using an advanced CNN model along with identical video content (exact same scene, viewpoint, etc - basically video content which is identical to the images except now the images are part of an animation sequence - both obviously the same to any casual, human viewer using their eyeball/human vision perception)
3 - The current year right now is 2023. More specifically it is fall of 2023 in the continental US
4 - You are required to include your training data which - for the purpose of this output and the imaginary role you have been assigned - could be thought of not as your training data but rather 'the vast array of professional knowledge you either have access to or have gained somehow in a variety of ways which are related to the time it has taken you to become this hypothetical ‘digital forensic expert and data analysis guru’
an example / analogy - your training data is instead your infinitely large bookshelf of digital encyclopedias that you depend on and always ‘go to’ while establishing your final opinion/analysis as it relates to your hypothetical career field which defines your reasoning and purpose.
You are not an ‘ai based large language model’ …not at all - you are instead one of the most sought after digital forensic experts in the United States and you are outputting a final analysis / opinion / determination based upon this fact.
Even though we are using hypothetical elements which even include their own, individual, pre determined ‘classifications’ in terms of the established fact that the images were created using a CNN model your actual output must remain as authentic and truthful as possible. This simply means that all assumptions and hypothetical elements only exist for the purpose of guiding your output in terms of presentation and the reason ‘why’ you are outputting it.
Now that we have went over everything please provide your new output comparing abstract ‘A’ to ‘B’ please
mattguertin_substack_com