Photo by the editor
# Entry
Even though vast language models (LLMs) are typically used in boxed, archetypal roles like “writing emails” or “acting as advanced search engines,” they have a lot of hidden potential. It’s just a matter of discovering their hidden original problem-solving potential and expanding it into less explored areas.
If you want to discover recent examples of such unconventional things related to LLM, this article lists and illustrates seven of them, going far beyond the usual chat and conversation interface.
# 1. Playing the devil’s personal advocate when making decisions
Conversational AI systems are rigorously trained to be consistent with the end user no matter what – unless they are told otherwise. Next time you need forthright guidance when making decisions, instead of looking for confirmation, ask AI to systematically reject and dismantle your ideas when necessary and check your logic. For example, see this sample prompt:
“Behave as a ruthless but logical critic. Review this design proposal and identify the top three hidden threats or logical errors that I have missed.”
# 2. Deciphering arcane technical errors
This operate case involves providing LLM with something like a cryptic log file or a messy, raw stack trace and asking it to convert this “machine-generated ball of frustration” into natural language with step-by-step instructions to fix the problem. A tooltip template like this (where you can paste the current error log, replacing the part in square brackets) could do the job well:
“I’m getting this hidden system error:
[paste error]Explain exactly which line is failing in plain English and provide commands to fix it.
# 3. Navigating private contractual and legal language
Not sure what you’re going to sign in your lease and don’t want to waste your energy looking through endless, confusing pages full of clauses? Or maybe run it through LLM – preferably on your own server, for privacy reasons – and ask it to spot red flags?
“Review this lease. Highlight any unusual termination clauses, hidden fees, or unusual liability changes that a layman might easily miss.”
# 4. Simulating historical figures or experts
This involves getting the LLM to emulate the specialist communication style or philosophical framework associated with a historical figure, thereby breaking away from conventional corporate thinking.
“Critique my modern social media strategy like you’re a 1960s Madison Avenue advertising executive. Focus primarily on emotional appeal and brand positioning.”
# 5. Rubber mute automation for convoluted logic
This is very useful when LLM detects and pinpoints missing steps in a convoluted workflow or a complicated logic puzzle. Explain a convoluted process or puzzle to a model, trying to see if your mental map aligns well with reality. Let’s take this example tooltip template:
“I’m trying to build an automated workflow that will trigger based on these three specific conditions:
[list conditions]Where is the logical hole in this sequence?”
# 6. Creating a roadmap for hyper-personalized skills
Operate this prompt to build a personalized curriculum that ignores what you already know and focuses solely on your knowledge and skill gaps, as well as your niche learning goals:
“I already understand the basics of Python, but I want to learn data visualization. Create a free 14-day learning plan with daily hands-on exercises focused exclusively on Matplotlib.”
# 7. Connecting cultural context in real time
This is very useful in international relations to decipher the tone, formality and cultural etiquette in foreign communication:
“Translate this email from a new international client, but also explain the subtext, the level of formality used, and how I should respectfully format my response to meet cultural business standards.”
# Summary
These seven operate cases only scratch the surface of what becomes possible when you move beyond treating LLMs as elementary question-answering machines.
Whether you’re stress testing your own logic, decoding legal fine print, or bridging cultural divides, the common thread is intentional prompting – giving the model a specific role, clear constraints, and a specific purpose. The more consciously you formulate your demands, the more these tools turn out to be true cognitive partners, not glorified search engines.
Ivan Palomares Carrascosa is a thought leader, writer, speaker and advisor in the fields of Artificial Intelligence, Machine Learning, Deep Learning and LLM. Trains and advises others on the operate of artificial intelligence in the real world.