Ah, we’ve finally arrived at generative AI. Yes, the title of this post was a very lame play on The Who’s “My Generation.” If you’re new here, welcome! You might want to read the first and second posts in this series before you jump into this one.

“Tell us what ChatGPT is already!!!”

I feel like I clickbaited you into reading two full articles before getting you to this, but I felt that it was best to ease into this topic.

ChatGPT is a generative AI program which was released by the company OpenAI in November 2022 and has been in the news ever since. ChatGPT has made headlines this last year largely around concerns of academic plagiarism, legal malpractice, and privacy.

So… what is it? It’s a chatbot that uses neural language processing and generation to follow prompts and generate new content. [Pause for translation.]

A chatbot is a computer program which is designed to talk to a human user in a way that mimics human conversation. GPT stands for “generative pre-trained transformers;” and chat is, well, just that - an AI chatbot. You can “talk” to the chatbot, and it generates text based on your request or “prompt.” Much like your at-home smart speaker, chatbots and smart speakers use neural language processing and generation to “speak like humans” even though they are computers.

Neural language generation: the process by which machines turn structured data or speech into text or speech outputs that humans can understand.

Neural language processing: A machine’s ability to perform conversational tasks, understand and recognize what is being said to it, and respond intelligibly.

Prompt: Direct, explicit instructions given to an AI tool to generate a specific outcome. You can enter a prompt conversationally, and you can make the prompt as vague or specific as you want.

“Is it like Google or Ask Jeeves”?

OMG, thanks for reminding me of Ask Jeeves. I loved that tool. If you didn’t grow up in the 1990s, perhaps you’ve forgotten the true delight of typing questions in human-speak into the search bar on Ask.com. This website launched in 1997 and sold for $1.5m in 2001. Jeeves was a bald, snappily-dressed valet who served as the visual representation of your personal assistant on the World Wide Web.

To answer your question, yes and no.

Yes: similarly to using Ask.com, you can type your request or prompt in plain English.

No: Ask.com was a character search engine. It analyzed the characters you typed in and generated a list of websites that might relate. It did not translate the intent behind your question using neural language processing. This is where ChatGPT differs.

“I don’t really get it…”

Yeah, I didn’t either until I tried it. So let’s see this tool in action.

I went on ChatGPT and typed in the prompt: “Write me a one-paragraph summary of Alexander Hamilton’s impact on the US economy.” The tool would within seconds generate what you’ve asked for. See a screenshot below of when I entered this exact prompt into ChatGPT on November 8, 2023.

So how does this machine know how to answer you? Don’t worry - it’s NOT sentient and replacing human knowledge. ChatGPT is a large language model form of artificial intelligence. The app is “scraping” or searching the internet broadly for information, processing it, and compiling it.

Scraping: Capturing data from the internet for use in generative AI.

It’s piecing together the most commonly found pieces of information and forming a paragraph that meets your request within seconds. You or I could accomplish the same thing by conducting internet research, but obviously we are limited to human speed, rather than computer speed.

Large language model: This is a form of algorithm (specifically called a “deep learning algorithm”) that is trained on vastly large datasets. By being trained on such large datasets, these tools learn to translate and generate text or other content.

How does the app know what to write? It doesn’t, really - it’s simply studying language patterns across ALL the articles, biographies, and found information about Alexander Hamilton it’s scraped off the internet and then regurgitated that information in brief.

What else can AI do?

Another example that I can share with you from personal experience: AI-generated images. I wanted to upgrade my professional headshots but didn’t want to A) leave my house or B) have to book a photography slot. Last time I got headshots, I paid $70 and wasn’t thrilled with what I got for it. I used a tool called StudioShotAI and got 60 AI-generated images for $21.

So how does this tool work? I uploaded 11 photos of myself in various lighting and angles, and the tool analyzed similar characteristics of these photos. From there, the tool forms a digital composite of the characteristics and generates an assortment of realistic images.

An actual former headshot of mine (not AI-generated)

An AI-generated headshot of me, based on photos I uploaded

The software is not perfect. Some of these did NOT look like me, while others looked remarkably realistic. Interestingly, the software puts you in more professionally-appropriate attire, hairstyles, and jewelry.

The ethics of generative AI

We’re going to talk in our next piece about the ethics of using AI in the workplace, the liabilities it can create, and how AI can augment the work of human beings. In the lens of creative uses (generating content, art, etc.), it’s an interesting ethical debate.

This is an interesting interview from a PhD student at MIT about the intersection and coexistence between art and generative AI. As postdoctoral student Ziv Epstein shares in that interview,

“When photography was invented, some painters said it was ‘the end of art’. But instead it ended up being its own medium and eventually liberated painting from realism, giving rise to Impressionism and the modern art movement.”

Technology and innovation are unavoidable. But we must adapt our societal expectations and frameworks to work with new technology.

Takeaways:

This is our third piece in a series about AI. Generative AI tools - namely, ChatGPT - have taken headlines by storm over the past year. Generative AI is able to use neural language processing and generation to communicate in plain language with users. The tool analyzes data in large language models to generate new content, such as written summaries or even images and videos. Generative AI has sparked ethical debate over how this counters human creativity and has the potential to disrupt creative and academic sectors.

AI Series #3: “Talkin’ ‘bout my gener-a-tive (AI)”