ChatGPT, Bing, Claude, Bard… the AI-based chatbot war has already begun. It can be assumed that this will take a long time, because the big technology companies have probably decided to bet really big in this field and therefore they will not give up at the first difficulties or at possible signs that the market may start to decline. be saturated from a certain point.

Nor can we expect them to respond to an open letter signed by big names in science and technology asking for a six-month “pause” until the necessary regulations are in place (and yes, indeed they are necessary, as the potential of artificial intelligence is extraordinary but poorly managed can lead to really complicated situations). Gaining an upper hand in this market that aims to be so disruptive, may become a golden calfwith all the negative connotations associated with it.

Right now, when many chatbots are in various stages of testing, their creators are competing to give them more features, increase visibility, and of course, to enrich them on the basis of a continuous training process, with which they know more about everything every day. This is necessary to prevent wrong answers to user queries, and also greatly reduces the risk of hallucinations, a common problem with AI that we will discuss in depth shortly.

Be that as it may, technology companies need to feed their training processes with data, whatever its origin. However, a Google used responses from ChatGPT to train Bard, according to allegations from a former company employee, as we can read in The Information (behind the paywall). The complaint was authored by Jacob Devlin, an artificial intelligence engineer who worked at Google during the trial and is now an employee of OpenAI. The accusation is quite clear, according to the article mentioned:

«Devlin resigned after he shared concerns with Pichai, Dean and other senior executives that Bard’s team, which received help from Brain employees, was training its machine learning model using OpenAI’s ChatGPT data. Specifically, Devlin believed that Bard’s team appeared to rely heavily on information from ShareGPT, a website where people post conversations they’ve had with ChatGPT.»

This accusation, which was rejected by Google As we can read in The Verge, this is somewhat worrying, because as we have already said on more than one occasion, ChatGPT is not infallible. So relying on their answers to train another AI model certainly seems like a rather unwise bet. So much so that, to be honest, I find it hard to believe that a team of professionals could have acted this way, even more so given what’s at stake with how Bard works. But it is true that Google had to step on the accelerator due to pressure from Microsoft, so it is not completely excluded that at some point someone decided to take a shortcut.