ChatGPT: Who’s Doing the Talking Here?

It’s no overstatement to say that ChatGPT has taken the internet by storm. Little wonder: when an algorithm capable of discussing any topic goes viral, reaching one million users in less than a week after its launch, we have to ask ourselves what we are doing, and because its based on an algorithm, what we are creating here.

What is the future of ChatGPT? Virtually everybody who tries it out and asks it to write something or answer questions sees it doing something interesting in their field of activity. But in practice, all ChatGPT does is compose conversations based on subsets of tagged information in a database that it uses as pieces, without worrying too much about their veracity or accuracy. In fact, while maintaining a chat with complete grammatical correctness, semantic coherence and even memory of previous conversations with a machine still impresses many, the reality is that the results, in many cases, leave much to be desired in terms of their level of depth or veracity.

If we were to evaluate many of the ChatGPT responses, we can see fundamentally what any high school student with a good level of language proficiency, and with a search engine and some time and skill, would write, able to turn their writing into something that looks coherent, but unable to apply critical judgment. With time and training, we can possibly get to the point where we could reach Ph.D. level in the relevant area. But we still have the same problem: how to ensure that the information used to construct answers is correct and appropriate. Or more importantly, how to prevent it from feeding back to itself previous answers produced by itself or by other similar algorithms.

The question, of course, is what now? Some argue that this kind of conversational interface is perfect for so-called informational searches, when users seek to learn more about a given topic or answer a question, which are one of the three types of queries typically defined in search engines — along with navigational ones, such as “take me to this page,” and transactional ones, such as “buy a smartphone” or “travel to the Maldives.” It is precisely these types of searches that have evolved the most over time and account for the bulk of a search engine’s value-added activity (we generally search for information many more times than we buy stuff).

Seen in this light, could ChatGPT be the future of search engines, or will it disrupt the search business to the point of threatening Google itself, that remains cautious on the use of this technology? That’s not an easy question to answer, because it starts from a very controversial assumption: that of feeding the algorithm’s database with correctly curated information to prevent it from giving wrong answers. And faced with this task, we must necessarily ask ourselves, as Google did some years ago when it proposed its Knowledge Based Trust (KBT), something as complex as what is truth, who defines it, or how to maintain the integrity of this page, with all its cultural differences properly considered, in the face of more and more information.

On the other hand, and given ChatGPT’s astonishing output in just five days, it is worth asking whether we are not at the end of the period in which human knowledge was susceptible to being searched and delivered instantaneously, a period that began with the invention of search engines, and that ends with the development of large conversational models using natural language capable of regurgitating information continually. In fact, we may already be asking ourselves the same problem that teachers like myself are asking: how to differentiate answers coming from a human — a student — based on research, writing and conclusions, versus the contamination generated by a machine at work all the time. How will experts in a subject, in some time, differentiate real contributions from those produced by an algorithm?

Regardless of what OpenAI intends to do with ChatGPT, we must ask ourselves not only what the possible uses of such a technology are, but also what we are helping build with the intense training process to which we have been subjecting the algorithm for more than a week. And above all, what the consequences might be.

And no, I did not write this article using ChatGPT.

—

This post was previously published on MEDIUM.COM.

***

You Might Also Like These From The Good Men Project

Compliments Men Want to Hear More Often

Relationships Aren’t Easy, But They’re Worth It

The One Thing Men Want More Than Sex

..A Man’s Kiss Tells You Everything

—

Photo credit: iStock.com