Users who confide in ChatGPT regarding work-related complaints or political leanings may regret it, according to an expert in artificial intelligence.
Mike Wooldridge, an AI professor at Oxford University, asserts that it is “extremely imprudent” to divulge private information or engage in intimate conversations with a chatbot, given that such disclosures aid in the training of subsequent iterations.
Additionally, he states that users should not anticipate an impartial response to their comments because the technology “reveals what you desire to hear.”
The topic of artificial intelligence is the focus of Wooldridge’s Royal Institution Christmas lectures this year. The organization states that he will examine the “major challenges confronting AI research and debunk the myths surrounding the true operation of this revolutionary technology.”
Among the subjects he will address are the inner workings of chatbots and how one can instruct a machine to perform language translation. In addition, he will address the looming question regarding AI: can it ever be genuinely human-like?
AI Understanding: Myths, Realities, and Cautionary Tales
Wooldridge stated that although humans were hardwired to search for consciousness in artificial intelligence, doing so was fruitless. He stated that AI “has no empathy.” It lacks any capacity for sympathy.
He further stated, “That is in no way what the technology is accomplishing, and more importantly, it has never encountered anything.” “The technology’s sole purpose is to attempt to convey your desired message; that is essentially its design.”
“It is prudent to assume that any input entered into ChatGPT will be directly fed into subsequent iterations of ChatGPT,” he cautioned. Furthermore, in the event that one comes to the realization that they have divulged an excessive amount of information to ChatGPT, retractions are essentially non-viable. According to Wooldridge, retrieving data that has been inputted into an AI model is virtually unattainable due to the nature of AI models.
An official representative from OpenAI, the company that developed ChatGPT, stated: “We implemented the feature to disable communication history in April. Our model training and improvement processes will not incorporate discussions initiated with chat history disabled.
Throughout the lecture series, Wooldridge will be accompanied by AI industry titans. According to the Royal Institution, he will also present “a variety of robot companions who will illustrate the capabilities—as well as the limitations—of contemporary robots.”
Michael Faraday established the Christmas lectures in 1825 at the Royal Institution in London with the intention of educating and entertaining young people about science. Having debuted in 1936, they currently stand as the most ancient science-oriented television series.
Nobel laureates William and Lawrence Bragg, Sir David Attenborough, Carl Sagan, and Dame Nancy Rothwell have all delivered lectures.