Branded Dangerous: AI-Chatbots

By Professor Dan Remenyi

We are being told that we are living in the most exciting times. And this is evidenced by the billions of words which have been written or spoken about AI. We do not seem to be able to let a day pass without a new announcement, revelation or debate directed at this technology. In the USA we have the suppliers of AI technology rushing to a Senate hearing in Washington to ask the Government to impose legislation on their activities, something that has never happened before. The UK Prime Minister flies to the USA to discuss with the President a number of things, one of which is AI and its control.

And what was the trigger for this? The release of OpenAI’s product called ChatGPT last November. So what is all the fuss about? ChatGPT is an AI-Chatbot which can understand colloquial language and respond in equally colloquial terms either through a keyboard or by voice.  It can hold a conversation with a human on almost any subject and supply answers to all sorts of questions. The AI-Chatbot can write reports, stories, poetry, and songs. And it can do this in a large number of languages. This has never been achieved before and it is on several levels quite remarkable. There are numerous applications for this new technology, many of which promise to help us create a better life for many people.

But there is a downside. Firstly, the AI-Chatbot makes mistakes. It can misinterpret a human question or remark. The AI-Chatbot has had to be trained and its training can be imperfect. When the AI-Chatbot makes a bad mistake, and they have made some really big ones, it is said that it is hallucinating. AI-Chatbots can in a number of ways exaggerate and be quite secretive. There are even reports that AI-Chatbots may have actually lied. There is the worry that some people might put too much faith in the replies of an AI-Chatbot and there is also the concern that some people might become addicted to conversing with this device.

And some of the AI-Chatbot producers have claimed that these machines have trained themselves on subjects which were not suggested by the manufacturers. This is perhaps the biggest worry. Some commentators are suggesting that AI-Chatbots are showing “sparks” of what is called Artificial General Intelligence (AGI). Having AGI means that the technology would be autonomous i.e. independent of humans and would continue developing itself until it is more knowledgeable and much smarter than any human.

Yuval Noah Harari, the Israeli public intellectual says that in no time these AI machines will be so smart that the difference between their intelligence and human intelligence will be similar as the difference between human intelligence and the intelligence of frogs. He then points out that this is highly dangerous. It is suggested that somehow these machines will take over control of the world’s affairs. And Harari correctly says that we humans do not have much respect for frogs. Could this happen? It is unwise to ever say “never”. But to my mind it is unlikely that anything like this could develop in the next few decades.

Nonetheless the AI-Chatbots are a remarkably powerful technology, and they could be used for nefarious purposes including misinformation, political manipulation and fraud to mention only three.

This subject is fully explored in Professor Dan Remenyi’s new book called Branded Dangerous: AI-Chatbots, available here, which will be available paper and soft copy from late July this year in the ACIL Bookshop.