ChatGPT, the smart echo machine
Paul Vauterin
CTO @ BioLizard / on the exciting crossroad between software, data & life sciences
I'm sure that, just like me, you recently have been exposed to posts about ChatGPT. Most of them go as follows: "I asked ChatGPT about X, and look at how smart, correct and to the point the response is. This is disruptive!". That response typically is a reasonably well written and moderately coherent echo of content and thoughts that are widely spread online.
But an echo, no matter how sophisticated, does not add anything new.
Sometimes these posts come from people whose jobs include text content creation such as marketeers and copywriters. And the next thought often is "this will profoundly change the nature of my job!". Which immediately makes me think: do you hold your job in such a low esteem? Don't you think that you can do better than that? Isn't your ambition to come up with unique and valuable new input, something no one ever came up with before?
领英推荐
It gets even more amusing when people ask ChatGPT about the prospects of itself or AI in general. You invariably get the echos of all the hypes people have published. That does not seem to prevent people from attaching some kind of distinct value to the fact that ChatGPT is reproducing them.
It reminds me a bit of how people were impressed by 19th-century automata.
To be clear, I do think that ChatGPT is indeed a hugely impressive engineering accomplishment. It's amazing what text extrapolation, deeply trained on a huge body of existing copy, aiming to be indistinguishable from human-produced text, can do. And no doubt it will have many good and bad usages. Future will tell about the balance, but one source of concern is: what if published content is increasingly created by tools like ChatGPT, and that content is used to train the next generation tools? Increasingly distorted echos of echos.
But most people seem to miss what to me is the most interesting and thought-provoking aspect. By showing us what can be achieved by a mechanistic approach like deeply trained text extrapolation, we are forced to rethink what is real intelligence and creativity consists of. It turns out that some stuff of which we thought it demanded intelligence and creativity, such as writing a mundane piece of text or poem, does not. In that way, tools like ChatGPT are holding us an inconvenient, but very interesting mirror. That mirror helps us refine our thinking about the true distinguishing aspects of our human intelligence.
One of the best posts I've seen so far about this Paul! Echo is the word alright. To me, Chatgtp and the like expose an uncomfortable truth which was hidden in plain sight for a long time: Humans are to a large degree mechanical,?reacting on a set of stimuli towards a predictable outcome, copying output from others, processing it in a path of least resistance. And we like being cultivated as such. If anyone would doubt that, just watch how people reacted at any aspect of the world cup. Or elections. To me, a part of what humans makes humans is the ability to have an intimate knowledge about a certain job, to 'genuinely care'. Cleaning staff will never be fully replaced by even the most expensive vacuum robots. Somehow we all know why. A software developer will never be replaced by chatgtp, nor will a teacher, nor a translator by online translation services. Not if they care and are able to stop and critically assess the job and its result as a whole. That is why I think its of such a great importance to re-invent people, in and outside the workplace, as wielders of tools and not duplicating them.
Guiding companies to become more cyber resilient by combining my extensive experience with adoption of recent AI based solutions
1 年Great insights!
Strategic Intelligence | CS&IO | Founder of Trensition
1 年Amen