Factiverse is a company that deals with fact-checking. But now it goes a step further. She received EUR 1 million from investors and wants to create a tool that will indicate artificial intelligence hallucinations.
Factiverse wants to fix AI
The problem with AI is that it is not uncommon for it to hallucinate. ChatGPT doesn’t know how to answer your question? He will give an answer anyway, but… a made up one. This is what blocks the adoption of new technology.
Factiverse wants to help. This is a Norwegian company that has tools based on machine learning. It uses them to verify content. Now the system will be used to improve AI. Or more precisely: getting rid of hallucinations.
Why is it so important? No, it’s not just about improving technology on principle. AI in its current form does not inspire confidence. Yes, it’s about hallucinations again. Example? The case of a lawyer from New York was widely reported. During the hearing, he cited what later turned out to be false court precedents. However, he didn’t do it on purpose. ChatGPT gave them to him.
However, the above is just one man’s mistake. CNET had to correct as many as 41 out of 77 articles prepared by AI. There were a lot of mistakes in the texts.
Now let’s look at the raw data. Rangaraya Medical College in India conducted an experiment. Scientists asked the OpenAI bot various questions and then verified the answers. As it turned out? Of the 178 AI “statements”, as many as 28 had no references in Google. In other words, the artificial intelligence made up the answers on the spot, i.e. it lied.
Will Factiverse help solve this problem? The company has been building its content analysis models since 2019. There is, for example, Factiverse AI on the market, which already helps check whether AI is not lying to us.
Not only Factiverse
Interestingly, AI hallucinations are creating… a new market. The RagaAI startup is also working on a mechanism to improve artificial intelligence. This company was founded by Gaurav Agarwal, a former employee of Nvidia Corp. and Texas Instruments Inc. The company received USD 4.7 million from investors to build a model for combating hallucinations.
If the problem of hallucinations can be solved, it will be a real breakthrough. Without this, no one will fully trust AI! However, if the mission of Factiverse or RagaAI succeeds, the continued existence of, for example, Google or Wikipedia may be at risk. There will also be a change in how we use the Internet. Instead of the websites we have just indicated, we will ask chatbots for information.
