By: Gabriel Fernández Chief Product Officer at Neuronic

Last year, I started teaching virtual classes in a course offered at CESTE, the International Business School of Spain. During a class on crafting prompts for use in Generative Artificial Intelligence, a surprising incident occurred that made me look like a beginner.

After the theoretical part, I wanted to put the concepts to the test through a simple example. I introduced a well-structured prompt into a new ChatGPT conversation to have it draft an article reporting on modern trends in renewable energy generation. In the video of the talk “Unraveling Language: The Revolution of LLMs in the Era of Generative AI and Innovative Chatbots,” which I presented at the GeneXus Meeting – GX30, you’ll be able to see the complete prompt:

So, with people connected watching attentively, ChatGPT’s response left me dumbfounded. It had a childish tone, a variable I didn’t include in the instruction. I tried again, and the same thing happened. It used phrases like “energy is like a super powerful magic that we use to turn on lights and make fun things work.” I had no choice but to apologize to my students and input the prompt into Google’s Bard, where it worked perfectly.

Later, I wanted to confirm why ChatGPT was responding that way. I entered the custom instructions, and in the box that asks, “How would you like ChatGPT to respond?” someone had given the following instruction: AS IF YOU WERE A 10-YEAR-OLD GIRL. I remembered that the night before, my niece Samira had been using my computer, so I deduced what had happened.

Ultimately, this highlights that generative AI is permeating our lives, and like my niece, what we are doing is altering the behavior of AI through instructions. A few years ago, one of my mentors told me, “Gabriel, the problem with companies nowadays is that they perceive technology as a module or as a component of the business, and they have to understand that technology is the business itself.”

But the concept of technology has now changed due to the significant advancements in LLM. Nowadays, we could say that AI is the business itself.

That’s why at Neuronic, we develop conversational agents based on AI and intentions that we can implement on multiple platforms and management systems using semantic search and the use of vector databases.

The journey taken by our team was full of challenges and learning experiences to optimize our product.

In the lecture, we summarized it into four fundamental points: writing a well-structured prompt, ensuring sensitive customer data, making it a reliable product, and ensuring easy implementation.

Lastly, I want to highlight what our CEO Germán Rampo presented at the beginning of the lecture, as all the progress made in recent years is due in part to the great technical ability of our team and in part to the philosophy of our company: positively influencing the lives of the people who make up the team. Keeping the team motivated and happy leads to better production and a smoother workflow.

Happiness equals progress.

Neoronic Team in GeneXus 30

You may also be interested in reading:

The 30 Most Viewed GeneXus Talks

Curiosities of the First GeneXus Meeting (+ photos and video)

37 talks about software and Artificial Intelligence