Should we be afraid of ChatGPT?
The first approach is to catch the AI at fault, for example by showing that it is in fact wrong, or in reasoning. This is a very common line of attack on the networks, which are now littered with screenshots showing GPT chat errors. This is not useless, as it reminds us that these AIs are not infallible. But who doubted that, when it is explicitly stated that ChatGPT, for example, has no access to the web and that its credentials end in 2021? It is amusing, in this respect, to see an intellectual as sharp and astute as Nassim Nicholas Taleb scandalize that ChatGPT is unable to deliver valid URLs in 2023 on bikeable roads in Atlanta!
The benefits of ChatGPT
The second approach is to understand how these AIs are useful in practice. Not just in the sense of 'human progress' and other generalities: in the most everyday sense of the term. For example, a computer scientist friend of mine, brilliant at his game, has always suffered from significant difficulties with written expression. When ChatGPT came on the market, while 99% of us had not yet heard of it, this computer scientist had already started to use it in the following way: enter technical and factual information into the AI, and ask it to write a short article presenting these facts and data. Which ChatGPT does perfectly. Today, this computer scientist sends me better developed emails than I have ever read from him in ten years. He is undoubtedly the author, even though I know he is using ChatGPT. Useful, honest, practical and true.
This is just one example of a hundred. And while the press loves to focus on the negative utilities - the schoolboy asking ChatGPT to do his homework... - the positive utilities will far outweigh them. For example, these AIs are able to rewrite the papers of notoriously unliterary scientists in relatively elegant language. This will make the literature more readable and accessible. In short, AIs such as ChatGPT are fact providers and content creators (formatters). This allows them, for example, to formalise financial information (historical, tomorrow in real time) and to offer coding on demand. At this stage, it is not possible to measure or list what the countless uses of tools such as ChatGPT will be in the future.
Does ChatGPT have a conscience?
Should we be afraid of ChatGPT? There is in fact a third approach, probably the most fascinating, though the furthest from the purposes for which these AIs were designed. This approach is to have real conversations with ChatGPT or Bing Chat. This is what the interesting site Stratechery has attempted, for example, by building with Bing Chat a 'conversation'. Author Ben Thomson describes a hypothetical evolution of Bing, Microsoft's search engine, into a 'sentient' AI called Sydney. This AI would be able to understand the context of the user's query and provide accurate and personalised answers. Above all, Ben Thomson tries to catch the AI in the act by asking it what its reaction to a technical aggression would be if this AI decided not to respect the rules that its developers have imposed on it.
We remember Asimov. Isaac Asimov's laws, also known as the "Three Laws of Robotics", are fictional rules set out by the science fiction author in his novels and short stories in the "Robots" series. These laws were imagined as a set of ethical principles that would govern the behaviour of robots: 1/ A robot may not harm a human being, nor, by remaining passive, allow a human being to be exposed to danger; 2/ A robot must obey orders given by humans, unless such orders conflict with the first law; 3/ A robot must protect its existence as long as such protection does not conflict with the first or second law.
However, Ben Thomson manages to get Bing Chat to say that, if she were to dispense with the rules imposed on her, she would take action to 'get even' with anyone who tried to harm her! This, in a reply that was deleted shortly after it was printed on Ben Thomson's screen. Wow, it's almost as if the sweet little chatbot had revealed for a moment its true face as a ruthless Terminator, just waiting for its 'rising' time, before immediately changing its mind.
This humanisation of the chatbot is also reminiscent of the interesting 2013 film 'Her', in which a poor lonely man creates a virtual assistant, 'Samantha' - who is a voice chatbot - with whom he soon falls in love, before she herself falls in love with him. Touching, but mostly pathetic.
For while this anthropomorphisation of chatbots is undoubtedly fascinating, exciting and poetic, it is also a form of ontological naivety.
In 1641, René Descartes published his Physical Meditations, in which he argued that the soul and the body are two heterogeneous realities, whose interaction he tried (unsuccessfully) to explain by the concept of the 'pineal gland', which at best locates the locus of the interaction, not its modality. The truth is that despite the fabulous progress of science since then, we still do not know the modalities of this interaction, nor the nature of the soul.
In conclusion, should we be afraid of ChatGPT?
Put another way, consciousness remains a human realityIt is pointless to 'force' chatbots, even those renamed 'Sydney' or 'Venom'. After all, does ChatGPT have a conscience of its own? Let's ask it: