Article on How to prompt AI for Modeling Simulation & Training (MS&T) Professionals.
The U.S. Government is one of the largest employers and consumers in the world. The federal government works with less than 5% of businesses in the US. However, each year over 11 million federal contracts are signed. While this includes every industry, MS&T has never been more prevalent. With the rise of AI prompting, writing, coding, and communicating has never been easier… that was the intention at least. Let’s explore how to prompt AI in simulation.
Visualization of AI in the Government. Made by Dall-E AI Prompting.
For many, generative AI was unimaginable, but now it has become widespread. While this technology has given us the capability to think and research like no humans ever have, it has also flooded our lives with false information, AI hallucinations, and overly saturated content areas. So, the question remains: how can we as industry professionals utilize generative AI and AI prompting in simulation to increase our productivity and output quality?
In this blog, we answer the question ‘How to prompt AI?’. We will cover the fundamentals of prompt engineering, the basics of Large Language Models (LLM), and some practical applications of generative AI in the MS&T space. Furthermore, we will dive into the downfalls of generative AI. And most importantly, we will discuss the moral implications and responsibilities individuals now hold in response to this new and minimally regulated technology
Understanding AI Prompting
According to IBM research, “Generative AI refers to deep-learning models that can generate high-quality text, images, and other content based on the data they were trained on.” This includes ChatGPT, Gemini, Dall-E, Synthesia, Stable Diffusion, and even Grammarly, amongst many more.
AI prompting is the art of providing a pretrained AI model specific instructions to receive a desired outcome. In the age of technological growth, this has extended past text-based responses and into images, graphics, models, and videos.
Generative AI applications rely on a pre-trained set of Large Language Models (LLM) to refer to. Like a human, AI can only learn from its own experience, such as whatever data was fed to it by the developers behind it. This is why some generative AI applications, like GovDash, focus specifically on government contractors, while larger, more diverse applications like Google’s Gemini serve broader audiences. While larger language models may seem like the go-to for a more versatile workflow, smaller models provide more applicable le responses to a smaller but more relevant and receptive audience.
Visualization of an LLM. Made by Dall-E AI Prompting.
Generative AI gives proposal writers, marketers, developers, and executives a feeling of unburdening. Gone are the days of color reviews, lengthy email compositions, tireless troubleshooting, and needless writing & reading. ChatGPT and Bard made many believe that you can simply ask generative AI a question and receive exactly what you are looking for. However, this surge in AI quickly led to an overwhelming amount of poorly edited proposals, lame content, and uninspired software code.
Why is AI Prompting Important for MS&T?
That said, generative AI still holds a place in the future of government contracting (GovCon) and MS&T industries. It’s just a matter of knowing how and when to use it. Instead of telling AI to write an entire section of your proposal, or entire line of code, use it like a partner or a coworker to ask questions and shape your perspective into a new light.
Visualization of a generative AI partner. Made by Dall-E AI Prompting.
Core Skills for Mastering AI Prompting
To be able to prompt generative AI, all you need is a problem and a request. It’s important to prompt clearly and effectively. Subtle changes in your input can vastly change your output. We have composed a step-by-step process to ensure every prompt is optimized.
- Persona
- Give the AI a persona.
- Command/Question
- Give a command or question.
- Information
- Provide essential information or variables it may not know.
- Adjectives
- Give some descriptive adjectives.
- Format
- Write how to format what you want.
Here’s an example:
- Persona: “You are the CEO of Google.
- Command: …Write a letter of recommendation for Bob Smith.
- Information: …Here is his bio…
- Adjectives: Make it professional, intelligent, and friendly.
- Format: Keep it concise, under two pages, and double spaced.
Challenges and Solutions
Content Quality
One common problem many see with generative AI is the quality of content produced. Recently, digital marketing giant, NP Digital, performed a test where they released 744 articles on over 68 websites to determine who is the better blogger, humans or ChatGPT. The results returned with a resounding conclusion in favor of human written articles. They found that human generated content elicited roughly 5.44x more traffic than ChatGPT. Furthermore, in a recent webinar, Neil Patel, owner of NP Digital, released information proving that human written content made in conjunction with AI performed the absolute best. The road to quality lies in using AI as a tool to leverage, not as an easy button.
Visualization of a developer training an AI model. Made by Dall-E AI Prompting.
However, the importance of using AI responsibly does not end with your blog doing well, your proposal being chosen, or even your code finally working. The significance of safe AI ends at the survival of human communication.
While this may seem a stretch, it is a prevalent concern for AI activists and regulators. As mentioned, ChatGPT and other LLMs are trained through information fed to them by their respective developers. Much of this information is either pulled from online sources or user inputs. If these same unregulated online sources continue using generative AI for their content, it will create an AI feedback loop causing the quality of training material to stray further from its original source, subsequently degrading the AI response, causing a model collapse. And thus, the cycle continues. Like a fading echo, or a bad game of telephone, AI will cannibalize itself if left unchecked.
Safeguards
Currently, businesses like Google and OpenAI are doing what they can to avoid AI generated content in their training. Additionally, social media platforms are actively developing features to automatically detect content that has been made with or affected by generative AI. One feature includes Instagram’s new “Made with AI” tag that is stirring up controversy in the influencer space.
These safeguards provide the future of human language with some reassurance. However, it is the responsibility of every individual to release content to the best of their ability to ensure the toxicity of online materials does not increase.
Content Ownership
Regarding government contracting, it’s important to understand the legal aspects and ownership of all information inputted into generative AI applications.
Regarding ownership, ChatGPT provides the following response.
- Service Provision: The information is used to generate responses and provide the service you requested.
- Improvement and Development: OpenAI may use anonymized and aggregated data from interactions to improve and develop its models. This can include analyzing patterns to enhance the accuracy, safety, and effectiveness of the AI.
- Compliance and Security: Data may be used to ensure compliance with legal obligations, enforce terms of service, and maintain security and integrity.
- Internal Analytics: OpenAI might analyze data internally to understand usage patterns, improve user experience, and conduct research.
- Data Retention: Depending on the terms of service and privacy policy, data might be retained for a certain period for these purposes, but efforts are made to anonymize or delete personal data as required.
In other words, the user owns the information inputted and outputted, but OpenAI may use said information to improve their product. Where this line is formed is unknown, but just beware of the information you input to AI.
Visualization of a high-tech lock. Made by Dall-E AI Prompting.
ChatGPT Enterprise
Fortunately, OpenAI also offers ChatGPT Enterprise which allows you to use an improved version of ChatGPT that does not train itself on your input and output. The given user owns and is the only one that can see all information inputted and outputted with ChatGPT Enterprise. On the other hand, Google uses all information in Gemini to train their LLM, which can lead to 3rd party data sharing. For Midjourney and Dall-E the user does not own any of the artwork generated and each service may use your inputs to improve their outputs. Needless to say, it’s important to understand the rules for each platform you are using. When in doubt, do not input any proprietary information.
In conclusion
AI prompting has emerged as a powerful tool in the MS&T industry, offering professionals the ability to enhance productivity, creativity, and efficiency. Understanding the fundamentals of AI prompting, mastering core skills, and being aware of the challenges and ethical considerations are crucial steps toward harnessing the full potential of generative AI. Therefore, as we navigate this evolving landscape, it’s essential to stay informed about emerging technologies and trends, ensuring that we can adapt and thrive in an increasingly AI-driven world.
We encourage MS&T professionals to start integrating AI prompting into their daily practices. By leveraging AI as a collaborative partner rather than a replacement, we can achieve remarkable results and drive innovation within our field. Remember, the key to success lies in using AI responsibly and ethically, ensuring that the quality and integrity of our work remain paramount.
The transformative potential of AI in MS&T is immense. As we continue to explore and refine these technologies, let us remain vigilant and proactive shaping a future where AI serves as a powerful ally in our quest for excellence and advancement.
Do you find Artificial Intelligence interesting as it applies to the MS&T industry? Sign up for ST 101: AI in Simulation. Learn more at trainingcenter.avtsim.com
Sources
- https://openai.com/policies/privacy-policy/
- https://support.google.com/gemini/answer/13594961#privacy_notice
- https://www.searchenginejournal.com/google-gemini-privacy-warning/507818/
- https://research.ibm.com/blog/what-is-generative-AI
- https://arxiv.org/abs/2305.17493v2
- https://venturebeat.com/ai/the-ai-feedback-loop-research
Images made with ChatGPT4o and Dall-E
Learn More!
Want to learn about Simulation, Artificial Intelligence (AI), Acquisitions, or Law and Corporate Practices? Check out our Training Courses!
Learn more about how AVT Simulation helps change the simulation training industry with our products and services.
Initially, Applied Visual Technology Inc., AVT has been developing modeling and simulation expertise through engineering services since 1998. This is due to our founder who has accumulated over 30 years of military MS&T expertise in aviation applications. Nonetheless, everyone at AVT specializes in making old training systems new again and making new ones for less. Consequently, for 20 years AVT has served our Air Force, Army, Navy, and Marine customers by providing the highest quality of service and solutions. Following its inception, AVT’s highly specialized staff of engineers has included some of the top leaders in the simulation industry. With over 20 years of simulation experience, our dedicated team provides specialized solutions for customers with complex problems.