Why AI Invents Answers (With Screenshots)
Unfortunately, far too many businesses (and individuals) have rushed to adopt AI without realizing that AI is not only often incorrect, but it will also invent answers if none can be found. Businesses that adopt AI into their business model without realizing this may find themselves in legal trouble in the future if they do not proceed with caution.
AI invents answers because it is not programmed to give “no answer found” responses. Additionally, if only a small amount of information can be found, it is programmed to fill the gaps on its own, even if it needs to fabricate answers. Read on to learn more about why AI invents answers and some screenshots of how AI hallucinates.
Why Does AI Make Things Up?
When AI companies first launched, they promoted AI as a way to quickly dig through large amounts of information. In their rush to convince companies to adopt their AI machines, they had to come up with a solution for when no answer can be found. Whether they realize it or not, the solution was that AI fills the gaps with whatever it can find, including information that isn’t true or doesn’t exist.
In more conclusive terms, AI makes things up because it is a statistical pattern predictor that is programmed to continue to make predictions even when there isn’t enough of a pattern to glean from. Basically, AI hallucinates because that’s how it was built, and it doesn’t have the capabilities to do its job when it has a lack of inputs.
Now this doesn’t mean that AI can’t be used for things like sorting customer queries, or tasks like spell-checking an email, but it does mean that companies who provide legal advice cannot (and should not) be using AI for research or anything which may need to be presented in a court of law.
How to Stop AI From Making Things Up
During the course of writing this article, we came across several articles with tips for bettering your queries so AI doesn’t make things up. But unfortunately, we have bad news for you: the only way to stop AI from making things up is not to use AI. Based on our research, current AI returns incorrect queries 45% of the time.
Sadly, so many companies switched to using AI far too early, and so many articles on the web were generated by poor AI over the last two years. This means there is so much misinformation on a topic that even if enough data points do exist for AI to give you an answer without hallucinating, it may, in fact, be pulling from a previous hallucination that a different AI generated years ago.
Not only that, but because AI was trained on human data, which could also include errors and biases, there is no way to guarantee even the best AI information can be error-free.
Honestly, at this point, companies that represent individuals on a legal basis, which need guaranteed true information, should not use AI and should only rely on primary sources.
Proof that AI Makes Things Up
We know the pro-AI readers are already clutching their AI-generated pearls in response to this article, but we can prove to you that AI makes up things very easily.
The easiest way to catch AI in a lie is to ask for information about an obscure topic. As part of maintaining this blog, we often delve into more obscure topics and have frequently caught various AI machines lying to us in the response because we were searching for a topic with too little information.
We have included some of our screenshots below, but if you are knowledgeable about an obscure topic, we urge you to try it for yourself. It’s not as hard as you may think to catch AI in a hallucination.
Here is a query we asked Google, pairing our website address with the keyword CoinMarketCap. Any human follower of our blog would know that we use CoinMarketCap only to track coin rankings for our altcoin spotlights. CoinMarketCap unfortunately relies heavily on AI for its token profiles and often gives a lot of inaccurate information. Other than tracking the tokens, which AI does get right, we do not use CoinMarketCap for anything else. AI fills this void with incorrect information, insinuating we use the website to track gambling trends…which any human who has ever seen the website would know that the website has nothing to do with gambling trends, and no business in their right minds would use that website to track gaming trends.
This is also not the first time we have come across lies like this about our website. We’ve also seen it lie about the contents of other websites. Unfortunately, there doesn’t seem to be enough true responses in the cryptocurrency space for us to even consider using it.
Will AI Ever Stop Making Stuff Up?
Of course, this article would be incomplete without discussing the future. As we mentioned, AI technology is changing and growing daily, however, we doubt we will trust it any more in the future.
As we mentioned above, the world became too reliant on AI too soon, and as a result, the internet is currently flooded with rehashed information we don’t trust. Not only that, but you can never guarantee what information AI is pulling from. Even if you set for AI to pull from only primary sources, you never truly know when it will start to hallucinate and fill gaps on its own because of how it was built.
Again, we are not stating that all AI is bad, but we here at MintDice.com will not be using it to research or write articles for the foreseeable future due to concerns about made-up information, where the information is sourced from, and the legal use of information on a professional level. We urge other companies to review their AI use and ensure they can stand 100% behind the information they are presenting as factual and true before posting it, no matter the origin.
