Close

Communications.Management.Marketing

Truth Is Just Perception

The Dangers Of ChatGPT For Brands

It will foster the advent of a new user interface and new content dynamics.

The ChatGPT1 chatbot uses a data set comprising 300 billion words and 175 billion parameters. To simplify its operating mode, let’s say that it constantly tries to produce the text it has written so that it resembles what a person could write. To do this, it relies on the large language model (LLM) with which it was trained, i.e. the hundreds of billions of words written by humans that it has ingested – the words, not the humans. It then looks for terms that have the highest probability of following the words it has just written.

As such, ChatGPT’s activity consists of asking itself for each word it adds: Given the text I have, what should the next word be? What makes its operation fascinating is that it does not always choose the term that has the highest probability and that there is no mathematical rule for the choices it makes in that regard. This is why the same prompt submitted to ChatGPT several times can generate different answers.

In the end, the “magic” of ChatGPT is that it writes texts that make sense because they sound like the ones it finds in its training material. Incidentally, as scientist and entrepreneur Stephen Wolfram has pointed out, this signals that “human language (and the patterns of thinking behind it) are somehow simpler and more ‘law like’ in their structure than we thought.

ChatGPT attracted one million users in the first week after its launch and surpassed 100 million monthly active users in just two months, making it the fastest growing consumer application in history. Beyond its meteoric success with Internet users, ChatGPT is exciting because it embodies the advent of a new user interface that will bring unprecedented ease and intimacy to the human relationship with digital technologies. For a while, we thought that the next man-machine interface would be vocal – based on the uses made popular by Alexa, Siri, and others – or that it would involve augmented or virtual reality equipment. But innovation is rarely where we expect it.

ChatGPT’s revolutionary potential in terms of user interface was highlighted a few days ago by the release of its first plugins, small components that add functionality to an application software. ChatGPT partnered with eleven companies, including Expedia, Kayak, Instacart, OpenTable and Zapier. Thanks to those plugins, ChatGPT becomes a real personal assistant that can help you book your vacations, for example. You just have to ask: “I’m going to Chamonix for a long weekend. Book me a train from Paris to Saint-Gervais for Friday May 12th in the morning and a return train for May 15th in the afternoon. I would like to stay in a hotel near the Aiguille du Midi cable car for X euros per night maximum. I would also like you to book me a dinner for two each night in a different restaurant.” Travelers no longer need to go to the various booking sites involved to organize their vacation.

Incidentally, those plugins also allow ChatGPT to connect to the Internet and thus have real-time access to current events, whereas until now it was limited to the events included in the data it was trained on, which ended in 2021. This allows ChatGPT to offer up-to-date information, which significantly increases its capabilities. Gradually, ChatGPT is turning into a real platform, which some observers are already comparing with an App Store. As always with digital technology, disintermediation creates players that rule the world.

The seemingly irresistible emergence of ChatGPT naturally creates a significant threat to search engines, first and foremost Google Search, whose owners have failed, unlike Microsoft with Bing, to anticipate and take advantage of this new technology. But its disruptive potential is at least as important for communicators and marketers.

(CC) Evan Ezque

Said disruptive potential is twofold.

First, the website may no longer be a reference content vehicle for brands since chatbots will offer a much better user experience2. Just as companies had to adapt their content to the advent of search engines by implementing organic and paid optimization methods, and then to social networks with the obligatory shift to interactivity with their stakeholders, they will have to adapt to generative artificial intelligence.

The challenge in terms of user experience will be coupled with an even greater challenge in terms of content. Indeed, as we have seen in the example mentioned above, chatbots draw information from online sources in order to answer the queries submitted to them. They therefore represent a form of natural selection of the most relevant content. This selective process will be largely indifferent to paid approaches, even if, of course, models that do not yet exist will certainly be invented in the future in that field. The fact remains that, in the future, the basic unit of content preferred by Internet users will probably no longer be an article or a video from a brand, but a conversational thread with an artificial intelligence: If you liked the algorithmic curation of content carried out by social media, you will love the one performed by chatbots.

The danger for brands is all the more important as, already, only 5% of their content generates 90% of their interactions with their audience: The other 95% of their content at best generate indifference among their stakeholders and at worst irritation because they do not meet their expectations. To paraphrase Montesquieu, useless content weakens the necessary content. It is impossible to overestimate the work brands will have to conduct to adapt to this new ecosystem.

Fortunately, said ecosystem is not exclusively risky3 for them. For example, they will be able to take advantage of artificial intelligence to produce more attractive content on a large scale: It is estimated that 90% of all content posted online in 2025 will be created by algorithms. This will enable the advent of personalized mass marketing and communications: Each content could be customized to its target audience.

But this will not be enough to ensure brands succeed in this new environment: It seems to me that only their ability to promote strong viewpoints will allow them to differentiate themselves from the content that chatbots will curate.

1 Chat Generative Pretrained Transformer.

2 An even more centralized digital universe could emerge in the future… unless the dynamics of web3 finally manage to materialize on a large scale.

3 And I don’t even mention in this article its dangers in terms of disinformation and cybercrime.

Leave a comment

Your email address will not be published

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Go up

Logo created by HaGE via Crowdspring.com

Carousel pic credits : I Timmy, jbuhler, Jacynthroode, ktsimage, lastbeats, nu_andrei, United States Library of Congress.

Icon credits : Entypo