Belgium, (Brussels Morning Newspaper) ChatGPT is a generative artificial intelligence agent that is based on a large language model (LLM) and is able to convincingly emulate human discourse to the point of passing the Turing test (becoming indistinguishable from human sentience).
Access to ChatGPT is public (subject to free registration). It integrates with the Internet via a plug-in. Leading search engines such as Google and Bing have added it to their offerings, giving their users the distinct impression that it is just another way of providing reliable answers to their search queries.
ChatGPT is likely to dominate search engines soon for three reasons:Â
- 1. Its output is in the form of digestible, bit-size text capsules, eliminating the tedium of having to scroll through dozens of search results and having to click on the links;Â
- 2. It appeals to authority by expressly claiming to have access to billions of documents; andÂ
- 3. Text is always perceived as way more definitive than visuals or audio.
Should this transpire, it would portend an ominous scenario. ChatGPT gets its answers wrong more often than not and when it does not know the answer, it “hallucinates”: confabulates on the fly. In short: it lies very often and then grandiosely refuses to back down.
The makers of this monstrosity claim that it is in counterfactual error only “occasionally”. That is untrue. Even the most friendly research estimates are that it hallucinates about 20% of the time. The real figure is way higher.
Recently, Geoffrey Hinton, the AI pioneer, confirmed this risk posed by ChatGPT in a wide-ranging interview following his resignation from Google. He warned against imminently being swamped with fake information, false news, and images, and of being unable to tell true from false.Â
Moreover: phrase the same query differently and you are bound to obtain an utterly disparate response from ChatGPT!
I posed 55 factual questions about myself to ChatGPT. My questions revolved around facts, not opinions or controversies: where was I born, where do I reside, who is my sister, these kinds of basic data.
The correct answers to all my questions are easily found online in sources like Wikipedia, my own websites, interviews in the media, and social media. One click of a button is all it takes.
ChatGPT got 6 answers right, 12 answers partly right, and a whopping 37 answers disastrously wrong.
It was terrifying to behold how ChatGPT weaves complete detailed fabrications about my life, replete with names of people I have never even heard of and with wrong dates and places added to the mix to create an appearance of absolute conviction and authority!
This is way more dangerous than all the fake news, disinformation, and conspiracy theories combined because ChatGPT is erroneously perceived by the wider public as objective and factual – when it is neither, not by a long shot.
The EU needs to adopt urgent steps to stem this lurid tide before ChatGPT becomes an entrenched phenomenon, especially among users who are gullible, ill-educated, young, or conspiracy-minded:
- If the creators of ChatGPT continue to refuse to fess up to the abysmal rate of correct answers afforded by their prematurely unleashed contraption, they should be made amenable to defamation and libel laws;
- The makers of ChatGPT should be compelled to publish timely and comprehensive statistics about usage and veracity rates; and
- ChatGPT is an ongoing research project. It should be banned from the public sphere and from search engines.Â
More generally, the EU should tackle the emerging technologies of artificial intelligence and their ineluctable impacts on the job markets, education, activism, and the very social fabric. Legal and regulatory frameworks should be in place when the inevitable encounter between man and machine takes shape.
AI is a great promise. But it must be regarded with the same wariness that that we accord technologies like cloning or genome (gene) editing.
Rigorous regulation should prohibit any deployment of AI applications unless and until they have reached a level of stability, fidelity, and maturity tested in laboratories over many years in the equivalent of the rigorous clinical trials that we insist on in the pharmaceutical industry.
Opinions expressed in the op-ed section are solely those of the individual author and do not represent the official stance of our newspaper. We believe in providing a platform for a wide range of voices and perspectives, even those that may challenge or differ from our own. As always, we remain committed to providing our readers with high-quality, fair, and balanced journalism. Thank you for your continued support.Sincerely, The Brussels Morning Team
