To get the most out of artificial intelligence tools like ChatGPT or Google Bard, you need to know how to ask questions and queries in these programs that simulate human conversation and provide relevant contextual information. Engineers specializing in AI or 'instant engineers' explain how to achieve this.
Behind the texts, images and various types of content generated by artificial intelligence or artificial intelligence tools, from technical training and education, there is the development and testing work of systems where 'instant engineers' specialize. Immune Technology Institute.
AI tools such as ChatGPT or Bard are very good at simulating natural language and can be used to program 'chatbots' (computer programs that “talk” to the user) or produce texts or graphics with quality comparable to that achieved by human capabilities. AI expert Javier Reguenko at IMMUNE (https://immune.institute) explains.
But “these technologies don't learn or act autonomously, behind their functionality and the content they generate, the hard work of experts, 'instant engineers,' in shaping requests to generate AI response and results,” he adds. Reguenko.
A 'prompt', which means notification, suggestion, response or request in Spanish, is a message or text that helps express a request to an AI such as ChatGPT, Writesonic or Dall-e that creates content or maintains conversations. Interact with a user or a virtual assistant like Alexa so that the AI understands it and provides a better decision.
An 'instant engineer' or 'physical or request engineer' specializes in designing and deploying mechanisms for 'conversations' or conversational AI systems that allow effective and user-friendly interaction and communication, the U Center explains. -Tad (https: //u-tad.com).
This type of engineering involves figuring out the best way to ask things from the AI so that it delivers the best result, designing and developing commands, creating frequently asked questions (FAQs), and establishing dialogue. Development of rules and guidelines to improve “digital interaction between humans and machines”.
“The aim of these engineers is to learn to talk with artificial intelligence, through 'prompts', the instructions we give to the AI system to generate a response, like a language model,” explains expert Agustin Cuenca. Neurologically based associative learning.
That's why these experts' tips for learning how to chat with ChatGPT or Bard, and getting the most out of AI in general, can be very helpful:
Ask clearly and be specific
“The clearer and more defined the question, the easier it will be for an AI system to give you a useful and accurate answer. We should avoid vague questions and try to be as specific and detailed as possible,” explains Reguenco.
Provide background information
If the question or request to the AI is context-specific, it is important to provide relevant information about that physical or situational context. According to IMMUNE, this will help tools better understand what was requested or queried and provide a more accurate response.
An appropriate context for the conversation may include relevant information, previous references, or the user's expectations.
Break down complex questions
“When a complex question is asked, it is recommended to divide it into smaller parts, because this division makes it easier to process and answer each element of the question, avoiding confusion,” says Reguenko.
Specify the answer format
“When there are preferences in the format of the desired response (for example: a summary, a list of suggestions or a conversation description), this need should be indicated to the AI so that it can adapt to the needs. Information,” suggests the expert.
Check, correct, and try again
After receiving an answer from AI, Recuenco advises “it's a good idea to take some time to review it and, if necessary, provide additional clarifications or correct the question to get a more accurate answer, as this will help the system learn and allow you to be more precise next time.”
“Evaluating the quality and consistency of the AI's responses and providing constructive feedback helps improve its performance and avoid errors, as well as biases such as gender bias,” he adds.
Use simple language
Javier Recuenco recommends expressing yourself, communicating with AI using clear, precise and natural language, and avoiding confusing ambiguities, jargon or contradiction.
Try different shapes and styles
From IMMUNE they recommend “experimenting with different types of 'prompts' or methods, such as questions, statements, suggestions, or examples, to find the one that best suits the goal and tone of the conversation.”
An example of optimal communication
As a link to the previous suggestions, Recuenco provides an example of a query for tools like ChatGPT or Google Bard, which are well planned and structured to increase the quality of AI's response returns.
It reads: “What is artificial intelligence and what are its main applications?”
The query is “clear, precise, and natural, provides an appropriate context for the conversation, indicates the topic and type of response expected, and uses a question-and-answer format that facilitates understanding and organizing information.” Expert.
In addition, reliable and up-to-date sources supporting the veracity and relevance of the data will be taken into account, when responding or document for processing (in those instruments prepared) may be referred to 'promptly'. like ChatGPT 4), he concludes.
Ricardo Segura
EFE – Reports
