Did Jarvis make this up? It’s a question that many people ask themselves when they write something with Jarvis. In the world of fake news, it is more important than ever to verify your sources. So when you get a fact from Jarvis it is important to not take it at face value. So, how do you know if you can trust AI-generated content?
Jarvis knows a lot about writing articles and using the information on the internet. But it does not know anything about what is true, accurate, or factual so when you are working with Jarvis you need to fact check everything that it gives to you.
All facts and quotes generated by AI need to be fact-checked by a human. Most AI’s use GPT-3 as their model for generating content. This model was created by researchers at OpenAI and doesn’t actually know any facts. All of the information it uses is in a sense, “made up”.
How does Jarvis Know that?
While the facts that Jarvis uses are made up, the way it uses them is actually quite ingenious. It has a database of millions of articles and facts that are all tied together by an algorithm. The more information Jarvis reads on a topic such as artificial intelligence or cloud computing, the better it gets at writing about those topics in general.
In 2019 Open AI consumed about 10% of all of the content that was available on the internet. The more information it has, the better Jarvis gets at what it does and that is writing articles on certain topics, but he gets better not because he “knows” any facts, but he can see patterns that a human can not.
There are patterns to the written content on the internet talking about entrepreneurship and that is different from the patterns of content talking about cloud computing.
When you run a template, a command or press compose in the Long Form Assistant Jarvis sends information to the GPT-3 model for it to figure out what comes next in your pattern. If it spits out a fact or a quote that is because in the pattern of your writing compared to the pattern that it knows about on the internet, that is where a fact or a quote would be placed.
Sometimes it puts real facts and quotes in when there is a piece of information that is used frequently enough that it becomes a pattern. For instance when you say Zig Ziglar said, and press compose it might generate the following:
“if you think you can or if you think you can’t, either way, your right,”
that is because it makes sense in the pattern of writing about motivation. But it is not actually a Zig Ziglar quote, that is a Henry Ford quote, but it SOUNDS like something Zig Ziglar would say.
Treat Generated Facts as Placeholders
The best way to treat the facts that Jarvis generates is as placeholders. Jarvis knows that based on the pattern of writing on the internet a good quote should go in that spot.
So if there is a quote about motivation with Zig Ziglar’s name attached, look up what he really said and replace the Jarvis generated quote with a real quote.
When Jarvis writes you a Bio and it says that you are a “3 Time Winner of the National Podcast Award” and it’s not true, don’t leave it in there, but think back to what awards you have won.
If your bio says “Over 300 episodes” but there are only 25 or 30, don’t worry about it, fill it in with the correct information.
If you are working in the Long Form Assistant make sure that you replace any inaccurate information as soon as it shows up because part of the context that Jarvis sends back to the GPT-3 model are things that you have left in the Long form assistant. It really goes back to the lessons that our parents taught us we were kids, “When you tell one lie it leads to another” or the quote from Walter Scott, “Oh what a tangled web we weave when first we practice to deceive.”
By the way, I used Jarvis to generate the quote after I typed ‘Walter Scott,’ and it got it correct, it saw the pattern where I was talking about lies and then mentioned Walter Scott’s name and realized what quote I was looking for.
Is that Information real?
So how do you tell if a quote or fact is real or not? The easiest way for most content is to copy the phrase or statistic with the quotes around it and paste it into google. The quotes tell google that you are looking for that exact quote or data, and you can see if the information comes up on a website (or multiple websites) that can support the data. Sometimes you will get good information, and sometimes you don’t.
If it doesn’t come up, then it’s probably made up by Jarvis to fill in the pattern of your writing. At this point, you have a decision to make. Can you find a replacement quote, do you take attribution of the quote or do you scrap it entirely?
The best solution is to replace the quote with a real quote, but make sure you add attribution. If there isn’t a good replacement for your generated fill in then consider rephrasing that section so that you aren’t actually quoting anyone and use it as part of your article.
What if Jarvis doesn’t know about a topic?
If you are writing on a very niche topic or on something that is new since 2019 (for example COVID-19) Jarvis will struggle if you don’t give him very much information and will generate fake fact after fake fact, but they will sound amazing.
But just like how Jarvis’s false facts will compound, the same is true of actual facts. If you paste good information in at the top of the Long Form Assistant when you begin writing your blog post and then add to it, you will end up with blog posts that have good facts in them.
Jarvis will use that context and work that information into the information that he sends to GPT-3 in order to generate more good information.
So if you are writing about COVID-19 and don’t know much about it, paste in the first paragraph that is on Wikipedia onto your Long Form Assistant, and it will help Jarvis to generate more good information.
So if you are writing about something that is in the news, but isn’t super common knowledge, or is new since 2019 use a website like Wikipedia or IMDB as your source for facts so Jarvis can learn them too for each article. Jarvis will never remember that data, but he will use it in that specific article to generate more accurate information.
You are the Human
Remember that Jarvis is a tool, not a human.
You can use him to generate good content and facts for your blog posts, but you need to make sure he doesn’t lie or misinform your audience in order to do so. It’s easy enough to check the information yourself by searching it on Google as well if you want to double-check.
And that’s it! I hope you found this blog post helpful. If you think that Jarvis would be helpful to you for your content creation needs feel free to use my affiliate link, you will get 10,000 free credits and I will earn some money as well. That money helps me to continue to create amazing content for you. It also helps me to be able to keep producing these blog posts and videos on my YouTube channel