A Computer Science Expert on How Philadelphians Can Benefit from More A.I. in Their Everyday Life

In a short amount of time, people personally and professionally have embraced certain aspects of A.I. (it’s already common to use it as a helping hand with email body copy or research questions, for example). However, though A.I. has come a long way, the help we get may not always match our intent, and the information it provides isn’t always accurate. When that happens, instead of saving us time and trouble, A.I. can cost us more of it.

To help us tap into the true time-saving, efficiency-boosting power of A.I., we reached out for the expertise and insights of Yang Wang, Ph.D., professor of computer science and program coordinator for the master of science in artificial intelligence at La Salle University. Dr. Wang and a team of La Salle-based researchers and educators are leading innovative work across the major fields focused on A.I., from machine learning to large language models.

Recognizing the transformative impact of artificial intelligence across every industry, Dr. Wang has led efforts to ensure AI education is accessible to all students—regardless of their academic background. One key initiative is the course A.I. for All, designed to introduce students from diverse disciplines to the core concepts of A.I. and how it can be applied meaningfully in fields such as medicine, marketing, education, and the sciences.

“For those who want to explore A.I. more deeply or pursue a career in the field,” Dr. Wang explains, “we’ve also developed a master’s degree in artificial intelligence. This graduate program serves both students who aspire to become A.I. professionals and those who want to integrate advanced A.I. tools into their work, whether in data science, research, or industry-specific applications.” (You can learn more about the program here.)

That “A.I. for All” approach provides a starting point for Philadelphians ready to embrace A.I. In offering the following tips, Dr. Wang also provides an important understanding of how A.I. really works–which may just make you a little more comfortable, even optimistic, about a future where the tool isn’t mysterious, but a useful part of your day.

Use the Right Model for the Right Work

Have you ever found yourself in a scenario similar to this? You were amazed by the results when you used A.I. to create a poem, but were frazzled by bizarre destination suggestions when you asked it to create a simple vacation itinerary. If so, you’re not alone. That kind of experience demonstrates the first important lesson of artificial intelligence—it’s not one-size-fits-all.

Choosing which type of A.I. model to use is your important first step. The most prominent artificial intelligence platforms (including ChatGPT, the defining brand of the market) are known as large-language models, or LLMs. As Dr. Wang explains, in recent years, A.I. in general has taken a leap forward, thanks to increases in processing power, the amount of data available, and innovations in deep learning models. But LLMs are the models that accomplish the viral feats that have helped make those advancements visible to the public.

“They’re what we call killer applications,” Dr. Wang says, referring to a term that makes a particular technology or product a must-have. “LLMs are popular because they allow a regular user to directly interact with the A.I. system in human language on a regular basis.”

But if you’re a coder or a designer, you want to dig deeper.

“There are A.I. models that specialize in programming, in images, in audio and in speech,” Dr. Wang says, explaining that users should research which models are best for their particular task before asking A.I. to do any heavy-lifting.

It’s not only the broad types of artificial intelligence that should guide your approach to A.I. Even within the categories of A.I. models, different modes can be best suited for different functions. Recently, models like ChatGPT have been promoting the concept of “reasoning” models, which users can choose to switch to within their platform. Dr. Wang points out that this is more of a marketing phrase than reality.

“These ‘reasoning’ models, in fact, do not reason in any cognitive or logical sense,” Dr. Wang says. “What actually happens behind the scenes is that the model is fed with a massive amount of data, and then it digests those patterns so it can recognize them when you interact with it. It doesn’t understand your question at all—it uses probability to predict which sequences of words are most likely to correspond to your question.”

Still, the reasoning models are useful, even if they take a little longer to complete their assignment. They should be used for complex tasks because they’re designed to break them down into manageable steps and perform multiple passes at a single task before presenting you with a result. Just make sure to give them a single, intuitive command that allows the A.I. to break down your request itself.

On the other hand, for standard, non-reasoning models, the more guidance and the more specific steps, the better. That brings up what Dr. Wang explains is one of the key skills of A.I. and the focus of his course “A.I. for All”—“prompt engineering.”

Engage in Prompt Engineering

It turns out that how you phrase a request to your A.I., indeed, impact the results you get. That’s where prompt engineering, which refers to the way you phrase and structure your direction to A.I., comes into play.

“It’s best to use a calm and firm tone because the large majority of English is written in a regular, neutral tone,” Dr. Wang says. “So that gives you the largest pattern matching and the best possible answers.”

Your tone is just a small part of getting the most out of A.I. Dr. Wang breaks down for students the key aspects of successful prompt engineering.

“There are several specific factors I explain to students in particular,” Dr. Wang says.

For one, you want to provide the context of the request. Is this an email for a client or a coworker? Is it a report to your manager or a client? A.I. won’t know the situation unless you explicitly tell it.

This also applies to what the writing should sound like. A.I. can’t know the tone or style that you want without some guidance. And it turns out the easiest way for the A.I. to know that is through what Dr. Wang calls “role-playing.”

“If you tell the A.I. model to generate a response from a grandmother, then the response will really be aligned with that perspective, knowledge, tone and style,” Dr. Wang says.

The technique works because it bundles all the subtle information that creates someone’s writing or speaking voice–from diction to the rhythm of sentences to grammatical tendencies–into one key point (the provided role or persona). The A.I. can then check its resources for language linked to that data point, whether it’s a hipster blogger, an academic, or, say, a magazine writer.

Your last big decision is how much freedom to give the A.I., a factor referred to as “temperature” within the field of A.I. Want something more creative? Pare down the prompt and tell it to generate something out of the box. But for hyper-specific, clear needs such as a series of complicated calculations, give it detailed instructions without wiggle room.

Correct the A.I.’s Mistake

Even with the perfect prompt, A.I. behaves probabilistically, and that means trial-and-error will always be part of its use. The key is to make that process as efficient as possible.

A good first step after a poor result, one that Dr. Wang teaches students, is to ask the A.I. to explain how it arrived at what it produced.

“The A.I. will explain, so you can see whether there’s any mismatch,” Dr. Wang explains.

You can even ask the A.I. to “help you, help it.” Once you’ve identified the communication breakdown, ask the A.I. to create a prompt for itself that will circumvent the mistake.

“Just ask, ‘How should I revise my prompt if I want to make it more clear or I want to achieve a given goal?’ Then it will generate a new prompt for you,” Dr. Wang says.

It’s also important to remember that within each “chat,” the interaction with A.I. should get easier over time. The A.I. gains context from each interaction, learning more and more about what exactly it is you’re looking for. But if all else fails and you’ve gotten completely off track, try starting clean with a new chat.  

Be Careful–But Also Optimistic

Even armed with the right approach to A.I., you need to be wary of its limitations and risks, professionally and personally. A.I. is notorious for fabricating examples and generating misinformation, so everything it creates needs to be carefully checked. Dr. Wang points to a famous case wherein a lawyer cited case examples in court, which he had used A.I. to pull. None of the cases existed.

A.I. engines are also run by companies and organizations who own the information you give them–which means the information is not private or secure.

“In one famous example, a Samsung employee asked ChatGPT to review some of the company code,” Dr. Wang says. “And that led to a huge security issue, where the code could be discovered by other users.”

That means that a working knowledge of A.I. ethics and privacy issues is not only important in and of itself, but to your professional future.

On this point, Dr. Wang acknowledges the common public narrative around A.I.–that it’s set to remake industries across the economy. While some fields, such as those in computer science connected to artificial intelligence, will grow, others will likely decline. These include those based on more straightforward mathematical analysis like insurance appraising and even certain types of computer programming.

But that makes knowing how to use A.I. all the more important, even if you aren’t driving the field itself forward. Someone with expertise in A.I. who can bring that expertise into a non-A.I. field has a unique and valuable set of skills. This is what Dr. Wang is seeking to support with his program at La Salle.

“It helps you to be future-proof,” Dr. Wang says.

Harness the Power of A.I. for Your Career

Dr. Wang sees Philadelphia as particularly poised to benefit from the A.I. revolution. This is driven in large part by our high-level educational institutions and startup infrastructure that support a local economy driven by health care, telecommunications and technology, all areas that are set to grow rapidly with A.I.’s development.

So, with some simple steps, you don’t need to stress about the future of A.I. You can even enjoy its help in the present. Dr. Wang suggests exploring how your newfound expertise can boost everyday life in ways that you might not immediately think of, from asking it—carefully—to help with small legal queries to helping you create a hyper-personalized diet plan that accounts for the many intricacies of your own unique biology.

“As an educator, I want to empower people with A.I.,” Dr. Wang says. “My goal is to help others become more efficient and more creative, and to support them in opening new possibilities across different fields of work and life.”

Learn more about the Master of Science in Artificial Intelligence program at La Salle University at www.lasalle.edu/ai.

This is a paid partnership between La Salle University and Philadelphia Magazine