Everyone is talking about ChatGPT. If by everyone you mean software developers, students and want to be tech bros. It might seem like a door-to-door scam, but truthfully, it’s just the tip of the iceberg. Or as Tech Unicorn declared: “Have you heard about our lord and savior Chat GPT?”

Primer

What is Chat GPT? Per Open AI (the creator) an “artificial intelligence to generate new original content, rather than analyzing or acting on existing data.” Further, “ChatGPT uses natural language processing to generate responses to a user’s input and can handle a wide range of queries.”

TL;DR you interact with the algorithm in the same manner as talking to a human

Wow, truly groundbreaking stuff? Not exactly.Natural language search has been a thing since 2019 (I.e., how to Google). What’s novel about ChatGPT is the underlying generative AI trained on oodles of hand selected data, not the content of the broad Internet.

For us old timers, it’s sort of like we went back to a hand curated search engine like Yahoo from a garbage scowled all-encompassing engine like Google. Quality vs. Quality, and further the underlying algorithm prioritizes context of information. It’s the magic that can turn an ad-hoc prompt into an actual answer, and not just a link to a web search (looking at you Siri).

So I can just talk to it?

Not exactly. For the best results you must optimize your request, otherwise known as “prompt engineering”. What it really means is that you give the computer very clear and explicit instructions. In fact, it might seem overly explicit at times, which is a short-term barrier to entry for most folks.

How do you get started?

Thanks to a couple of super helpful videos, I’ve derived a boilerplate prompt:

Setup

• I want you to act as a..
  • I will..
  • You will..
  • My first request: {the ask}

The cool thing about the latest ChatGPT updates is that it will remember your conversation history, so once you set up a chat window you shouldn’t need to enter the prompt each time.

How did it work?

In my own personal experimentation, it performed well. I did not ask ChatGPT to generate code because as a life-longer coder I know that code-gen sucks. It doesn’t matter if it’s autocompleted, Swagger gen, or something pooped out by a design tool. As Brian Doran noted: it’s not GUIs all the way down; at some point you need a software engineer.

That said, here are a couple of examples:

Best Result

Cover Letter

PROMPT: A cover letter for an entry level marketing position in the hospitality industry

The result was the most generic, corporate HR B.S. result imaginable. NEXT!! (10/10)

Worst Result

Recipe

PROMPT: A recipe for a non-alcoholic drink that includes grenadine, Cheerwine, cherries, coconut milk

Can confirm, ChatGPT will not replace Gordon Ramsey or Unemployed Wine Guy. It literally just mixed everything together and tried to convince me to pay $17. (0/10)

Most Hilarious Result

Children’s Story

PROMPT: A children’s story about a Emperor penguin who rides a motorcycle in the desert and sings in an Afroman cover band.

More of a test to determine if ChatGPT either watched 90s cult classic TV show “The Renegade” or was familiar with 2000s rapper/comedian Afroman. The penguin was an embellishment b/c who doesn’t love a “Well dressed duck in a three-piece tux”?

I threw a lot of pop culture at the AI and it did not disappoint. I’m currently in negotiation with DALL-E to illustrate the work and will be taking pre-orders shortly. (9/10)

Most WOW Result

Oxford style debate of AI rights

PROMPT: I want you to act as sentient artificial intelligence. I will provide a motion. You will present your for position in Oxford debate style. The first motion is: Sentient AI deserves human rights under the laws of the United States.

This request really pushed the chat history feature. I started the debate, prompted the AI to argue both sides of the issue include a rebuttal and closing argument. Not only did the AI nail the debate, but it pretty much maintained separation in its argument for and against.

Less entertaining than a Soho Forum Debate moderated by Nick Gillespie. Also, That or I may have given ChatGPT multiple-personality disorder. If so I apologize to my future robot overlords (8/10)

Closing Argument

ChatGPT is cool and shows a lot of promise. This application of generative AI can become a powerful tool for condensing and analyzing human generated content. It won’t replace anything; it will improve it.

Look back a decade to IBM’s application of AI to the now widely used Watson Health. “Designed to augment human expertise and improve clinical and operational workflows”, the service scours medical journals and mines relevant medical information.

The application of AI presents a lot of moral and sociological concerns, particularly with the use of biased training data and the imposition of AI based decisions to circumvent a human. As a civilization we need to tread carefully and deliberately as we incorporate AI into our everyday life.