Ask an AI Insider
Q: What's all the hype about ChatGPT? And why is it so controversial among experts in the field?
You're probably already familiar with the media storm and public attention around ChatGPT, the generative text AI tool from OpenAI. It's been so exciting to see a LOT more people (100 million users in fact!) excited about the potential of AI and investigating how it might be used. If you're looking for ways people are using the new tool - from acting as an interview coach to a writing editor to a nutrition advisor - you can check out some of the great 'prompt directories' like 'Awesome ChatGPT Prompts' or this guide from AI Exchange.
However - there's a lot to be skeptical about, and reasons to use caution if you're looking to use chatGPT beyond playing around to create amusing (of not very good) poetry. Reaction from AI Industry leaders (Meta Chief AI Scientist Yann LeCun) and academics (Noam Chomsky, Gary Marcus) has been less enthusiastic that the public reception.
This isn't just sour grapes over OpenAI's success. Along with concerns about plagerism and copyright infringement, large language models are known to 'hallucinate' facts, sometimes when you least expect it. They aren't just 'confidently wrong' - they're deceptively convincing, missing nuances that are difficult to fact-check. Even experts at Google, knowing this weakness, accidentally botched the launch of their ChatGPT competitor Bard by promoting a conversation with factual errors. If Google can miss such a high-profile mistake - are you still confident you'll be able to check your own work?
With this in mind, a few guidelines for using ChatGPT are helpful:
- Leverage your own creativity before potentially watering it down by succumbing to confirmation bias. While using ChatGPT for brainstorming writing prompts, gift ideas or essay outlines can save time, it can ultimately stifle your own creativity by returning unoriginal ideas. Consider the dangers of using generative art in forensic sketching tools - are you wiping out your own most original ideas before they are fully formed?
- Consider the potential SEO impacts before copying and pasting any output for public consumption. Ross Dunn, SEO of Stepforth Marketingand host of the SEO 101 podcast, pointed out the issue in his latest episode: " From the school of hard knocks [...], it's very clear whenever something gets easy to do in SEO, it gets hammered down fast because there's going to be people who overuse it." Detecting AI-authored content is possible. Are you willing to do the editing necessary to 'trick' the currently-available detection technology? Or re-write once this technology improves?
- Stay paranoid about factuality. Many 'prompt engineers' and guides will warn you to check your work, but dismiss the difficulty of catching these errors. Consider the time it takes to meticulously fact-check all your work. And if someone tells you they have found the secret to get 100% correct answers and that you can automate content creation without a human review - they are trying to sell you something.
- Do explore these tools to learn their uses and limits. But don't get lost in rabbit-holes. Far too many people are investing hundreds of hours trying to design the right questions to get reliable answers - when this has been proven to not be possible. ChatGPT will just be the first of many tools that will help you automate mundane and mechanical tasks. There will be more, and it will become clearer how they really deliver value. It's worth learning the basics to get familiar, but this is NOT your last chance to leverage AI in your life or in your business. The AI race is on - but with each new innovation comes a chance to catch up.
What questions do you have about ChatGPT, generative AI, or leveraging these tools in your work and career? Email us and we'll be answering your most burning questions in future newsletters.