Director, Learning Design & Development| PMIAsheville, NC, United States
Validating and checking outputs is critical when working with AI systems like Generative AI. Such validation approaches may include establishing clear criteria, implementing strong testing protocols, and continuous refinement.
In your experience with AI, what are some best practices for ensuring the results you receive are accurate, relevant, and aligned with your original goals?
Be specific and give examples so that the AI has the best chance of having the accuracy, relevancy, and alignment that you are looking for. the AI is only as good as the prompt it is given. Saving Changes...
Many times if the output generated by AI is not what exactly I am looking for, I provide inputs to it to refine the output. Many times I could get the required output after few iterations. Saving Changes...
Claudio TantignoneSubgerente de Infraestructura y Arquitectura Tecnologica| Trenes Argentinos OperacionesCaba, Caba - Buenos Aires, Argentina
It is important to define the details of what is intended from the beginning, that is, if the scope is defined precisely, the AI ​​can give us the information we need.
If we are generalists and lack information, there will be errors that the AI ​​or anyone who helps us will make due to lack of information.
We must be detailed, structured, and concrete in the questions and expectations. And as we move forward, we must refine the details to achieve the desired result. Saving Changes...
Trever SimesIT Project Manager| ElwynHernando, Ms, United States
I hope this works so I can get my check Saving Changes...
Roger BrowneSr. Release Coordinator| Elevance HealthLindon, Ut, United States
"Clearly Define Goals: Start with a well-defined goal and ensure your prompts are specific and detailed to guide the AI accurately. Verify Sources: Cross-check the information provided by the AI with reliable and authoritative sources to ensure accuracy. Iterate and Refine: Continually refine your prompts based on the results you receive to improve relevance and alignment with your goals. Evaluate Context: Consider the context of the information provided and its applicability to your specific needs, making adjustments as necessary." Saving Changes...
This is an essential question. Regarding the LLMs, I believe most users, even is they use best practices like proper prompting technics, there is always the limiting factor of confidentiality and policies of each enterprise, limiting the details and examples that can be shared to receive the most accurate and helpful responses. Saving Changes...
Provide detailed instructions, use the RFT or CREATE format, and provide examples for your AI to leverage in it's response Saving Changes...
Nicholas BeaudoinDirector of AI Programs| Caltech (CTME)Pasadena, United States
Jumping in as a new member here. There is a tremendous amount of excitement in the AI dev world to see Gen AI as a software engineering toolkit. Hook up to API and integrate. But the traditional lifecycle that data scientists spent the past 10-15 years building out relies on the scientific method. It relies on becoming a subject matter expert (SME) of your data before you begin solutioning. We are seeing this disappear in the age of Gen AI and it hurts the accuracy and reliability of the models. The original question posed was how do we make reliable Gen AI solutions.
A great example is your naive RAG-based system. If you are looking to talk to your documents then creating a PDF reader for SEC filings is probably the wrong approach. Use something that you are truly an expert on. This is why we saw HR teams being the first departments that created value with Gen AI in industry. HR has a wealth of documents that only they understand; documents that are intrinsic to only that firm. When the Gen AI solution gave the wrong answer, HR professionals knew and could pass that on to the technical teams. The HR teams knew their data and were SMEs in the process! They could identify haphazard results and test the reliability while keeping a human in the loop! Saving Changes...
In my industry, it is essential to have the GenAI cite it's sources. In addition, I agree with the module to include as much detail as you can and make your requests and clear as possible. In addition I also wonder if it would be possible once GenAI has curated your product, to state your original goals and ask GenAI model if the product aligns with these goals? Saving Changes...
When using AI systems, it’s important to follow best practices to ensure the results are accurate, relevant, and aligned with your goals. Start by clearly defining your objectives and crafting specific, well-structured prompts to guide the AI effectively. Always validate the output by cross-checking with reliable sources or applying your own expertise to identify inaccuracies or biases. Provide feedback to the system if possible, refining prompts or parameters to improve results. Additionally, be mindful of the AI’s limitations and consider the context in which it’s being used, ensuring its outputs align with your broader objectives and ethical considerations. Saving Changes...
"Imagine if every Thursday your shoes exploded if you tied them the usual way. This happens to us all the time with computers, and nobody thinks of complaining."
PMI Team Member
