You have /5 articles left.
Sign up for a free account or log in.

Over the past year and a half, the advances in generative AI have been astounding. The impact on the way we conduct research, teach classes and develop new learning materials is nothing less than revolutionary. Yes, there are limitations, and yes, there are misfires along the way. The current versions of generative AI are not yet at a standard that the general public has come to expect from high tech today.

However, hallucinations are on the decline as OpenAI, Anthropic, Google and the other providers tweak their large learning models. It is a process of training the models to be more precise while not impinging on the creativity of responses. As Rahul Pradhan writes in Datanami, “AI models are only as effective and intelligent as the data set they’ve been trained on. Often, AI hallucinations can occur and are a result of a variety of factors, including overfitting, data quality and data sparsity.” Progress is seen in each succeeding generation of the LLMs.

Yet we must not simply sit by twiddling our thumbs while the rest of the world is adopting the technology and adapting to the current shortcomings so that we can implement the awesome capabilities as soon as possible. There is so much that we can experiment with, and accomplish, when it comes to the current models that are already available. Pradhan suggests, “Prompt engineering gives AI models additional context which can lead to fewer instances of hallucinations. This technique helps LLMs produce more accurate results because it feeds models highly descriptive prompts.”

We all can improve our prompts through practice and following suggestions from the generative AI apps themselves. For example, I asked Google Bard how I could improve my prompts, and it responded with multiple suggestions as well as a reading list that included the following from Google Cloud: “Best practices for prompt engineering.”

Go ahead and ask your favorite generative AI app how to improve your prompts—for example, Google Bard, ChatGPT, Claude 2 or another of the many apps. Submit some prompts, then ask, “How can I improve my prompts?” While you are at it, cut and paste your prompt into two more chat bots. Compare the results and compile the best solutions among the three.

To further improve and tailor responses to the inquirer, we are seeing developments in retrieval-augmented generation (RAG). This technique is beginning to enable the generative AI to take into account more current and detailed information to give enhanced context to meet the needs of the inquirer. Oracle Cloud Infrastructure technology content strategist Alan Zeichick describes the process:

“Today, in the early phases of RAG, the technology is being used to provide timely, accurate and contextual responses to queries. These use cases are appropriate to chat bots, email, text messaging and other conversational applications. In the future, possible directions for RAG technology would be to help generative AI take an appropriate action based on contextual information and user prompts … RAG might also be able to assist with more sophisticated lines of questioning. Today, generative AI might be able to tell an employee about the company’s tuition-reimbursement policy; RAG could add more contextual data to tell the employee which nearby schools have courses that fit into that policy and perhaps recommend programs that are suited to the employee’s jobs and previous training—maybe even help apply for those programs and initiate a reimbursement request.”

It is important to every discipline, every student, every faculty member, every staff member and every administrator that we become familiar with and adept at using generative AI. For students, this is the rising tide that touches virtually all careers. It is especially important in fields such as management, accountancy, psychology and a host of other fields that, in some cases, can anticipate a wholesale reconfiguration of position descriptions and declining demand for new hires.

HR departments are placing a high priority on hiring those applicants with documented portfolios demonstrating prompt engineering skills. Writing in The Wall Street Journal, Chip Cutter reported on the rapid rise in salaries offered to qualified applicants with extensive AI experience: “Salaries for AI roles vary based on the experience required and the company hiring. Total compensation, which typically includes bonuses and stock-based grants, can push overall pay much higher. A product manager position for a machine-learning platform at Netflix lists a total compensation of up to $900,000 annually.”

For faculty and staff members, and especially administrators, facility with chat bots will result in far greater efficiencies and even the ability to rapidly create detailed reports that, in the past, would have taken too long to develop and customize for the institution. Our abilities and competitiveness as individuals and institutions will increasingly be tied to the enhanced productivity enabled by generative AI.

During the holiday break, perhaps you can find time to develop your generative AI skills. If possible, run at least one prompt every day. Make sure when you run the prompt on one chat bot, you also run it on two other apps to compare the results. This will also help you spot hallucinations if they pop up. Over time, your skills will grow to become faster and more expert at using these emerging tools. It will become as customary as posting a Google search question today.

We need to begin to consider generative AI to be our personal assistant. Imagine that it is an actual assistant, albeit a little quirky, that stands by ready to help 24-7. This assistant has passed the interstate bar exam, posted near-perfect scores on the SAT and GRE, and on most bots, its knowledge is up-to-date within a couple of weeks. It is blazingly quick, thoroughly researching and posting an answer far faster than you could have even just typed the response. In most cases, you will find that you will need to further refine your original prompt with a few follow-up inquiries. The follow-up is part of the process of getting the response you are seeking.

Invest the time now to learn how to pose effective prompts and assess answers by comparing to other chat bots and sources. Continue to practice daily. This will pay off in your efficiency and effectiveness as the quality of generative AI applications improve. This will put you in a position to best serve your university, your students and your own interests as you grow with the technology.

Next Story

Written By