avatarsrgg6701

Summarize

Technology, ChatGPT

Knowledge, Truth, and Bias After ChatGPT

Like any powerful technology, chatGPT can make the world either a better or worse place.

Image by Susan Q Yin

For several months, the Internet has been obsessed with ChatGPT, and there is no end in sight. Events are moving so fast and on such a large scale that it’s time to talk about the dawn of a new era in the history of high technology.

In retrospect, it would not be surprising to see ChatGPT as one of the most disruptive technologies in the history of the Internet, if not the most disruptive. It literally changes the rules of the game for the entire WWW before our eyes: never before has a computer been able to communicate with a human being in a natural language for the latter and to “understand” the context of the conversation.

This way of interacting with the user is already a real dialogue as people understand it rather than communicating through an interface. It’s amazing. Suppose you are interested, for example, in how certain past events relate to today. In that case, you can take a trip back in time, talking about the details of those events with a knowledgeable personal guide. This, of course, applies to a host of other cases as well.

This whole new experience changes how we think about technology, allowing “them” to come much closer to becoming an extension of “us.” Even when ChatGPT makes a mistake, it is not upset or annoying because such a hiccup feels like a forgivable imperfection for a reasoned being rather than the dumb aloofness of a soulless machine. And even though it’s been months since ChatGPT appeared, it still feels like something fantastic.

But it’s not just about how it feels. It’s also about how our way of working with information is changing. The most important significance of the emergence of ChatGPT is its role as a catalyst for changing the information paradigm of modern society. Until now, this paradigm has been based on the following sequence:

  1. Problem definition
  2. Gathering data for analysis
  3. Converting data into information and
  4. Converting information into knowledge

Web search is a characteristic implementation of this paradigm, and working with a traditional search engine consists of the same steps. At the same time, ChatGPT allows us to see this process from a different perspective, taking the 2nd and 3rd steps outsourced. As perceived by the user, it does not collect data and does not transform it into information but immediately answers the question posed.

Of course, behind the scenes, a process similar to what is done in traditional search develops. The data is extracted from different sources, ranked, and combined. But all this is hidden and completed in a fraction of a second. Thus, for the user, the above chain is reduced to just two links:

  1. Problem definition and
  2. Acquiring knowledge

Once established as a new information paradigm, this transformation has the potential to affect all of modern society profoundly.

Of course, it’s not just about the user experience with ChatGPT but also how we acquire knowledge and perceive this process. Knowledge is the cornerstone of our civilization: the faster we accumulate it, the more advanced society becomes. One effect of the intensification of knowledge might be a consistent reduction in the need for individual mastery of purely technical skills.

Why would one spend years studying them if everyone could get them here and now? (It certainly wouldn’t look as spectacular as the iconic scene from “The Matrix,” where the knowledge of flying a helicopter was downloaded into Trinity’s brain in a few seconds, but the principle is pretty much the same.) Most of this knowledge may be moved from the human brain to the AI repository.

In this way, advanced AI can free humans from routine more than ever, allowing them to focus as much as possible on tasks that require creative thinking.

But technology is a double-edged sword. History shows that there are always those who will try to abuse it. ChatGPT is an excellent candidate for such uses. I will not dwell on its technical applications because something else seems most perilous.

Namely, it is the perfect tool for implicit persuasion. It does not impose the “right” answer on the interlocutor, but its possibilities to influence the latter’s opinion are potentially unlimited.

The very structure of the user experience with ChatGPT implies that its conclusion is accepted as true. Providing links to sources of information (which he’s not providing now) is unlikely to change anything.

People are notoriously prone to seeking not objective truth but confirmation of their beliefs, often, alas, denied by trivial prejudice. Which they, however, take as truth and insist on, including by violent means. This has always been the case throughout human history, and it is, unfortunately, an empirical pattern rather than a deviation in human behavior.

You can certainly see that people succumb to prejudice in many other situations, including using conventional search engines. Yes, this is true, but nevertheless, a regular search engine and ChatGPT represent two different situations.

In the case of a web search, the user is in an essentially pluralistic environment: having various sources of information indicates a choice, which influences one’s perception in one way or another.

In the case of one-on-one communication with ChatGPT, however, the same user finds himself in an echo chamber where he is given a single answer.

And this answer, on which all the user’s attention is focused, depends only on the algorithm behind the generated text and the data on which his “interlocutor” has been trained. The more satisfactory the latter’s answers are, the less critical the former becomes. Already, ChatGPT’s uncritical responses can be worked out very quickly, and the amount of time it takes to work out is likely to decrease as the technology develops.

In other words, ChatGPT is capable of being a very powerful means of indoctrination.

But even leaving such extreme cases aside, there is still plenty to worry about. What if the ChatGPT algorithm prioritizes user satisfaction over the objectivity of response? What if this algorithm is unprotected from illegal or directive interference by a third party?

After all, if we delegate control of stages 2 and 3 of the transition from data to knowledge to machines, who will be authorized to control these machines?

However, there is nothing fundamentally new in this question because as civilization develops, our reason becomes more and more abstracted from the processes it needs to control. Accordingly, the level of control that machines are charged with is also increasing. In many ways, this has become routine, but this does not mean that the issue is not severe enough.

On the contrary, because, firstly, with each turn of technological progress, the level of complexity of all this machinery increases, which becomes a problem in itself. And secondly, modern society still needs to compromise between the right of a private company to organize its business processes by its own interests and the right of citizens to be protected from manipulation of their consciousness.

But hardly anyone would argue that such a compromise is necessary. And no technology that affects man’s perception of the world by influencing his beliefs should remain a black box for everyone. Google, Facebook, and Twitter are textbook examples of the darker sides of high technology that emerge over time. The early years of these companies filled users with enthusiasm and excitement. Still, little by little, this attitude changed: the flipside of the tech giants’ generosity was the insecurity of user data and the exploitation of negative user emotions to ensure high clickbaiting.

It is very hoped that AI technology will serve people and not use them in the coming future, as it, unfortunately, happened in the era of “before ChatGPT.”

You may be interested in the following related topics:

Dear reader, subscribe to my updates and share your thoughts on our present and future with me!

Recommended from ReadMedium