Are you using generative AI correctly?

posted 6 min read

Years ago Elon Musk said that the best career for the future is not computer science, instead the world need people who can use the existing technologies correctly. Large Language Models (LLM) are a branch of Artificial Intelligence (AI), a lot of people now use LLM in their everyday life, the most common of them is ChatGPT, LLM are very useful tools which can save us a lot of time, effort, automate many things, and so on, is hard to find someone who don't know or had never listened about ChatGPT, but the question is, those millions people that use LLMs, are using them correctly?

What is a correct use of AI

For a better understanding of the definition, first lets discuss the main problems or possible wrong uses of LLM.
The philosophy of AI developers is create tools that help humans to automate repetitive tasks, facilitate data analysis, research, help humans in his everyday life work, always that such work is ethical, however every time you use a large language model, or any artificial intelligence tool you are using data from a database, and at the same time you are giving your data to the AI model, behind this model in most cases there are a lucrative company, which is saving your data into their databases for using it in the future for generating income, sometimes for improving the model and sometimes for selling it to third parties, but you may have asking yourself, what data I'm giving to them? The answer is the Prompt information, when analyzing data for you enterprise for example, you give this information to ChatGPT and ask him to retrieve an specific result, but know your information is in hand of other company database, and you not have nor the control of it, nor idea of what they will really do with your information, as an student you maybe are giving to ChatGPT your assignments so that he can solve it by you, and save time, get a better score, but you are nit learning, not practicing and in real life when you need this knowledge at work, you'll need to ask the Chat again and tow things happen, one is that you generated dependency and the second is that you given now maybe sensitive information about your company.

Another problem is the type of words you write in your prompt, is true that AI models are trained to be polite and kindly with humans, but when you write aggressive words into the prompt, those words are stored on the database, this can potentially give us two results, one is that the AI can in the future use those words to response your messages, the second is that maybe in a near future security agencies like CIA, FBI, can use your Chat history to evaluate your behavior, as well enterprises can use your history to analyze your capacity to work under pressure, and the care you have with sensitive data, it play a crucial role in the future of AI.

The next point is optimization & efficiency means, how fast and how effective are the Chat's answers you request to him, by fast I means the least prompts possible to get the answer you need, and by effective I means, how accurate is this answer.
Said that I conclude that correct use of AI is:
Getting fast and accurate response by the LLM, use those answers for generating knowledge and understanding, not dependency, managing your emotions and using polite communication with the Chat, and very important avoid sharing sensitive information with LLM.

How to use LLM effectively?

It depends on the goal, the first question you have to do yourself Is, what I am using AI for? These tools have a wide range of application and uses, lets discuss the most common of them in detail.

Data Analysis

ChatGPT for example is good performing complex operations of data analysis or mathematics, but when you need to analyze your business data or any sensitive data, it is a mistake to give your information to the bot, why? Because the company owner of the AI model will save your information on her database, and you don't have any idea what they are doing, or what they can do in the future with such data, what become an ethical issue, how to mitigate this problem?


Tip:
Instead of using ChatGPT or any other AI model for analyzing your data, create your own tools, Python is a programming language that have many packages like scikit-learn, pandas, numPy, scyPy, statsmodel, matploblib, and more, this programming language allow you to create your own tools, but they don't have a database which is saving your information, these tools are open source and the organization behind them, don't have access to what you are using Python for, it means you can create your own data analysis tools with python, without risk that somebody is watching your information

You don't need to be a programmer to do this, you can ask the AI to generate a code for analyzing and visualizing your sensitive data, but this time you are not giving sensitive information to the Chat, only your program will access your information, in conclusion your data keep private and secure.

Research

If you are making a research AI can be your ally, but you have to take care of the information, because some times the chat give obsolete information, always ask for link references so that you can validate the information.

Studying

With AI models it is easier to make a research, but it can lead to dependency, to avoid this, use AI for getting knowledge, and understanding, so that you can replicate this knowledge without needing the LLM.

Coding

When working with large system, generative AI will help you to generate pieces of code, I had proved with large code files and let me say, ChatGPT is bad with it, instead ask for specific pieces of your code and it will save you a lot of time and will multiply your productivity.

Tip: I recommend to avoid giving sensitive data about the project your created, in the future we don't know what this information will be used for.

Behavior

Be polite, act like if the model you are chatting with is another educated person, if you are polite with the AI, it will be polite with you, and your chat history will give a good impression about yourself, remember in a near future prompt engineering will become a fundamental and maybe necessary skill in your CV.

What about redaction?

If we want to redact an essay, report, article using generative AI, we have to give the necessary information of what we want, but not specific information about what we are doing, when I need to do it, I apply this technique: first ask to generative AI model for redacting a draft about lets X topic, with the specific and exact structure I want it or I need it, I use this draft as a template for replacing the AI information with my own information, if needed I leave the useful information, I saved a lot of time, I didn’t shared my data and my content is not plagiarized.

Ho to get fast and accurate answers?

This is a topic of prompt engineering, from my experience in the use of generative AI, I can say that first you have to be polite and specific, explain what you want? And How you want it? Give all the technical details so that the AI know how to show your information.

Why dependency is a problem?

All generative AI models are based on existing information, on the actual big data, if we become dependent of AI, we are becoming dependents of existing data and with serious problems for creating new unique data, meaning that all the new information will be based on the actual information and it reduce human creativity, the Novel Physics award to AI pioneers in 2024 confirm that human is using AI without boundaries, of course if have many benefits but if not used with responsibility it can become more a problem than a solution.

Conclusion

Generative AI are essential tools that when used responsibly help us save time, be more productive, but if we don't use them effectively as human specie we can encounter with problems in the future because of a lack of privacy and creativity, or become dependent of AI.

If you read this far, tweet to the author to show them you care. Tweet a Thanks

More Posts

Discover how to use AI writing tools without losing your authentic voice as a content creator.

Jimmy McBride - Oct 11

Deploy an AI journalist chatbot with Gradio and Supabase

Astra Bertelli - May 18

Is AI the Next Picasso? Exploring the Potential of Generative AI

Nwabueze Jude - Apr 30

Meet the .csv datasets and understand how they works on creating expert system Artificial Intelligence models

Elmer Urbina - Oct 19

Build a Telegram bot with Phi-3 and Qdrant and chat with your PDFs!

Astra Bertelli - May 9
chevron_left