Can you write a blog article about the differences between Bing Chat, Google Bard, and ChatGPT? Can you provide me with some factual information about their differences and how to best use each one?

That was what I asked Google Bard in a prompt and it returned a pretty decent answer.

Of course, I was curious and decided to ask the other AI’s for their thoughts as well and a similar answer was returned from each. 

Generative AI (of which large language models are a type of sub-model) are great at answering questions and can be a powerful tool for individuals who need to quickly ascertain understanding around a specific topic. However, as noted in the image above, generative AI is a model trained on available data from the internet and other database sources that can be used to train an AI. If you aren’t sure what “train an AI” means, just think of how you read a book, then comprehend and store the information that you read. That process is very similar to how generative AI models work. 

As such, generative AI are not able to develop new, original, and creative pieces of work. Anything you ask them will be taken from data and sources that they can access across the internet or from private servers. However, this should not dissuade you from using it, most of the information we utilize on a day to day basis are not new or creative. In fact, this article was going to tell you how to best use generative AI for your own purposes, but AI tools can explain that already – very likely by pulling it from documentation on generative AI. 

Of course, there are some added tips to get you up and running with generative AI. 

There are three generative AI models that can be used freely on the web: 

Bing Chat (requires Edge browser)

Google Bard (requires an account)

ChatGPT (requires and account)

There are others too, such as DeepMind, Cohere Labs Generate, and PolyAI. At the end of the day, all of these AI tools do the same thing: comb through large pieces of information to help you condense your tasks down to easier sizes. Knowing how to use generative AI is great, but any blog article written by an AI will come across as robotic and soulless (although… sometimes people can write this way as well). There are already a number of tools that can be used to detect AI responses. With that being said there are still ways to use AI productively while retaining your own creative flair behind it!

1. Be as Descriptive as Possible When Writing Prompts to an AI. 

The AI, in its response above, stated that you should be specific, but you should also be descriptive. For example, telling an AI to “Write me an essay about climate change” will produce a generic result that thousands of people will likely have already seen. Remember, AI can’t be creative so it’s taking all of its results from other bits and pieces of information that is plastered across the web. 

Instead as the AI to: “Write me an essay about climate change. Provide a description about the causes of climate change and how humans have been able to use instrumentation to show changes in global temperature. Also write a paragraph about increased vulnerabilities certain communities have faced due to climate change and why.” 

In the above prompt, we ask AI for a very specific essay, but we also ask it to give us detailed information on particular points within that essay. This helps us to keep the AI focused on a particular subject within the essay.

2. Always Check the Sources (References) the AI used.

Even when an AI returns a response, you should always ask it to provide references. The prompt could be something like:

“Can you provide me with a list of references you used to support the above statements.”

The AI should provide you with a list and here comes the hard part! You have to check those sources. It’s very possible the AI could just take the title of a text and use that to prove support for a claim, but the text behind the title says something entirely different. Reviewing the sources are still important and will always be!

3. Have some domain expertise in the subject you are asking AI about.

People have been using AI for everything from game development, software engineering, essay writing, and art generation. However, if you do not have the appropriate knowledge to understand the result you won’t really know how to apply it. For example, if you ask AI to provide you with some methods of making money via side hustles it can return a lot of results. However, if you asked it to teach you how to start working in those areas, you wouldn’t get clear instructions. 

You could ask it for follow up information, but even still, not having direct experience means you are merely accepting what generative AI says as truth – which can be very wrong (the internet is very often wrong about many things). It is important to seek out experts and ask them to share their experiences, wisdom, and guidance for any endeavor or subject that you are pursuing. 

4. Never Claim AI Work as Your Own

There are still a lot of blurred lines surrounding ownership of output from generative AI models. However, if all of the outputs by a model are taken from the works of other people across the internet, then you are very likely plagiarizing if you take output and claim it as your own work. 

This harkens back to point number 2. Always ask for sources and cite those sources – not the AI. It’s generally bad practice to cite wikipedia (and wikipedia cites the sources for their claims as well), so do not cite the model you used to generate a response. Look for the source behind the AI’s response and cite that or another source that supports the AI’s claim. 

5. Do not copy and paste content, modify it to fit your own tone.

We get it, some forms of writing can be pretty robotic (*cough, cough* scholarly texts *cough, cough*), but an AI’s prose is typically unemotional. You can’t feel the soul or the passion of the writing because the model is reciting facts (I’m sure this update is coming to all of the models in the future!!!). The best thing you can do is take content, paste it, look at it, and modify it to fit your style and language of writing. 

Authors have a way of communicating their emotions regarding a subject through the words, style, and syntax they use. It’s easy to call out AI writing, because it has none of this. 

Summary

Generative AI provides a new way for humans to quickly gather, synthesize, and communicate information. It is important to remember that much of the information provided by AI are taken from across the internet. Users need to verify the accuracy of the information provided by generative AI, modify it to sound human and fit their needs, cite legitimate sources for claims, and ensure they understand enough about the topic before making broad and authoritative statements on a subject based on the output from AI. 

Hopefully, this article provided some helpful ways you can improve your engagement with AI models in the future!

2 Comments

  1. All very valid and great points, Nick! I think you nailed it with the third point and it’s not something I’ve explicitly thought about, but having some domain knowledge definitely helps a person be more descriptive, as well as help the author add their own thoughts and emotions

Leave a Reply

Your email address will not be published. Required fields are marked *