By the end of last year, a week I was scrolling through the Internet as usual, and by the next one every single meme I received were ChatGPT answers. From people tricking it into think that it was the user, to answering 5 when asked “What is 2 + 2?”. A system that behaves this way cannot be said to be intelligent, surely.

ChatGPT was released at the very end of 2022, and it quickly became the fastest growing application ever. There is some sense of concern about it mainly due to its philosophical and ethical implications, but also because of some of its eerie responses.
So… what ought we do about it?


2 plus 2

ChatGPT’s core issue may be that it can “make up facts”, as was claimed by OpenAI’s CTO, which is a weird thing to say given that it is generally agreed on that “facts” are something that can be proven to have happened or that are the case.

The mistakes that ChatGPT makes are innocuous. It is just saying that 2+2 = 5. Or so one would believe. ChatGPT is not an amusement exercise, people use it to consult information, get help with homework, etc. And this has real consequences.

Take the case of Alexander Hanff, who got told by ChatGPT that he was dead. It not only claimed that Hanff was dead, but it curated a fake link to justify itself saying that. As Hanff himself points out, ChatGPT’s “mistakes” are real and could harm the reputation and opportunities of people, for if he was applying for a credit or a job, he would have been ruled out because he would be considered dead if the company was using ChatGPT to investigate its applicants.

Recently, a Go-playing AI could defeat top players, but would fail against an adversarial tactic. This is a clear blind-spot from the AI, for its training data did not contain such possibilities. Said example is merely a game, where no one would get hurt. But take the example of autonomous vehicles, where they can crash into objects, causing irreparable damage, or maybe even worse. The AI is not learning about the world in the way we do. Obvious tactics or scenarios are easily figured out in our mind, but not in the system’s model.


Idíõma

One could argue, as has already been, that AI tools like ChatGPT implicitly discovered that what once was considered the human quality, language, can be said to behave much more like laws governing the world. I am not opposed to this view, since it can replicate content that we have created in the past, or, what is the same, mimic it.

However, one must pay attention when one thinks about the concept of replicate (as Blade Runner would sure suggest). ChatGPT, by its very nature, mimics what it found, the statistically representative relationship in the text, represented in the utterance of the next word. There is no new content, or ideas, it is merely refurbishing what other humans already thought, wrote down, and shared it with the internet.

There are things that ChatGPT cannot just do, at least in the same level of proficiency as the others. Poetry is one of them, and I believe the reason is twofold: ChatGPT is not experiencing the world, like we humans do, it is only muttering combinations of what millions of humans reported about their experience of the world. And it is trained on the semantics of the text, not necessarily in the syntax or rules of specific texts (such as a haiku).

Language is the combination of words and the grammar that rules the way in which they are said. However, idíõma sheds light into the fact that it is a peculiarity of ourselves, this is, we pertain to ourselves insofar as we create our language and we use it for we are unique and replicating content should not be our goal.


Acroases

All of these things, plus the downfall of traditional social media, are what made this (I hope) consistent blog happen. My response to the question possed at the start of this first post is to do more: to produce our content that is truly ours.

And that means crafting more thought-out posts, the ones that are personal, curated to the readers of the internet that resonate with our own voice. I was heavily influenced by the POSSE (Publish Own Site, Syndicate Elsewhere) practice, referred by Cory Doctorow’s Pluralistic (which is a must-read).

In the midst of content that looks alike, self-replicating texts, one ought to create, publish and discuss ideas, however small or inconsequential, because your experience and your life is worth being examined.

Maybe being human is the fact that we have a spirit in the etymological sense of the word, because there is some breath inside us. A voice that is supposed to be used, and we have the tools to do so more than ever.
But maybe we are not and that is up for discussion, and that is precisely why this is acroases, because there also ought to be some listening to one another.