Software Development
AI

What do we expect from ChatGPT?

AI plagiarism tools, including OpenAI's AI Text Classifier, are starting to prevent passing ChatGPT answers as your own work. You're wasting its potential if you're trying to get it to do the work for you, anyway.

AI is becoming more human-like, but it's a far cry from human. (Photo by Pexels user Tara Winstead via a Creative Commons license)

This editorial article is a part of Big Tech + You Month 2023 in Technical.ly’s editorial calendar.

Open AI’s ChatGPT continues to enthrall and disappoint users.

Enthrall, because it’s ostensibly a bot that can write essays, poems and articles with a simple prompt. Disappoint, because the content it outputs isn’t that good, especially on your first prompt — and that’s if you can break through the virtual log jam and access the chat box in the first place.

A lot of users, especially those in industries where AI is likely being adopted, love to hate ChatGPT. It turns out boring copy, and it has a habit of confidently outputting inaccurate information.

And it’s true. But ChapGPT isn’t meant to write essays and articles for you.

Underscoring this is the fact that AI plagiarism tools are becoming accessible, including OpenAi’s own AI Text Classifier. This new tool detects AI text and is available for free. Like ChatGPT itself, it’s not perfect, sometimes misidentifying AI text as human text, especially in languages other than English.

I tried it out with one of my own articles and a couple of ChatGPT answers I’d saved. For mine, it detected that it was human-written (or “very unlikely” to be AI). For the AI, one was detected as “possibly” AI while another was classified as “likely” AI. There is also an “unclear” classification. Again, it’s not perfect.

Screenshot of AI Text Classifier classifying ChatGPT text as likely AI generated.

AI Text Classifier classifying ChatGPT text as likely AI generated. (Screenshot via Technical.ly)

Still, detecting AI is a growing industry, with platforms like GPTZero and Writer AI already making it harder for students to turn in papers generated by AI.

As for webmasters and clickbait websites? If they’re using AI as content, they’re breaking Google’s webmaster content rules, and will be penalized in the same way Google’s 2011 Panda Update penalized human content farms.

So then, what’s the point?

It’s not about doing the work for you. It’s fun to ask ChatGPT to write an article or a story or a song. Expecting it to produce human-quality content without a lot of human input isn’t realistic, but that doesn’t make it of no value.

There’s a learning curve

ChatGPT is a tool, like a camera. Most people have a camera in their pocket and know how to use it, but it takes creativity and skill to be a photographer.

The most direct comparison is an AI art generator like Open AI’s DALL-E 2. Like ChatGPT, it has its ethical issues, but one misconception is that people who use it to make art are just dropping a quick description and letting the bot do all of the creative work. DALL-E is evolving, but it’s not truly creative.

Your first try on an AI art generator is going to look terrible, nothing like the intricate results that use start images, multiple prompts and evolutions.

Whether you like AI art or not, making it look good is a skill.

What ChatGPT is good at

ChatGPT is designed to react and evolve as you talk to it — that’s what makes it different from the old AI automated customer service chat boxes we’ve interacted with for years.

If you talk to it enough, you can develop a “voice” for it — creating interactive AI characters is a fairly popular fan pastime using platforms like character.ai, where users chat with fictional characters (and sometimes “real people”) and play text adventure games (hi, Gritty). But it’s also used as a learning platform to practice communicating in a new language, conduct mock job interviewers and ask for advice. ChatGPT is more advanced, and can do all of those things; it’s all matter of prompt engineering.

ChatGPT isn’t a great writer, but it can clarify how things work, offer academic sources, simplify a choppy transcription, show you how to format a script or a resume, even help you come up with ideas. The creativity (and fact checking) come from you.

As an example of it demonstrating a format, here is an automated prompt in “radio ad” format letting users trying to access ChatGPT know that it remains overloaded:

[Narrator]

Welcome to ChatGPT, where AI technology meets conversation.

[OpenAI engineer 1]

We’re excited to introduce you to our new chatbot model, designed to interact in a conversational way.

[OpenAI engineer 2]

But as we launch ChatGPT, we want to let you know that we’re currently experiencing high traffic on our website.

[OpenAI engineer 3]

We apologize for any inconvenience and appreciate your patience as we work to make sure everyone has a chance to try out ChatGPT.

[OpenAI engineers, in unison]

So please check back soon!

[Narrator]

Thank you for choosing ChatGPT.

If you’re still trying to get into the popular platform, good luck to you.

Engagement

Join the conversation!

Find news, events, jobs and people who share your interests on Technical.ly's open community Slack

Trending

Delaware daily roundup: Delmarva Power vendor stats; DelDOT's $15M federal grant; 50 best companies to work for

Delaware daily roundup: Over 4,000 Black-owned businesses uncovered; Dover makes rising cities list; a push for online sports betting

Philly daily roundup: Women's health startup wins pitch; $204M for internet access; 'GamingWalls' for sports venues

Philly daily roundup: East Market coworking; Temple's $2.5M engineering donation; WITS spring summit

Technically Media