How the artificial intelligence of ChatGPT can impact the design of tomorrow’s content but not only.
Chances are, lately, you’ve heard about the ChatGPT phenomenon and all the stories that have started circulating about it. But what is this chatbot really? and above all: “is it really just a chatbot like all the ones we already know and use, for example, as a support within websites?”
Let’s start in order.
ChatGPT is what researchers and scientists in the field of artificial intelligence call a “large language model”, i.e. a neural network trained on information datasets (corpus) present online and capable of efficiently analyzing millions of different patterns of the way where people connect numbers, words and symbols giving them a meaning and in turn generating a final output that is understandable.
This is the real revolution of ChatGPT, this model, released in November 2022 by Open AI a non-profit research body in the field of AI, in fact, is not only able to understand the language with which you interact and generate answers specific questions about content or definitions, but can:
These are some of the capabilities of this model. In essence, we can consider this neural network as something decidedly superior to a normal chatbot and which will probably revolutionize the way people approach and interact on a digital level.
Tools like ChatGPT can be very useful for companies, and will be even more so when they are further optimized. The fields of application are varied and, if used correctly, they can help people incredibly, optimizing work times; in particular for all those tasks that require a high level of automation, but also, as we have seen, for those that require content production.
Some of the criticisms leveled at these tools are related to a potential replacement of people; however, the development of these neural networks is not yet such as to be able to give them (and perhaps never will be) some unique human abilities such as:
In fact, the logic of these models proceeds more by imitation than by actual learning and this leads them, in some cases, to make mistakes. So now let’s see what the limits of these technologies are.
The creation of content starting from requests or information provided is not always perfect and it happens that the model can invent some theories from scratch or incorrectly report some facts. This phenomenon is called by scientists as “hallucination effect”.
Part of this effect is due to the fact that the model learns by sifting through information taken from the network that may or may not be incorrect. This is one of the main problems that AI is facing right now. A second huge problem of ChatGPT (and of all AI in general) is that, for the same reason mentioned above, it is very susceptible to bias, thus risking providing the user with answers, reasoning and theories containing prejudices or beliefs do not verify.
An example is described in this tweet, where the model created a function for a programming code with the aim of defining a good scientist and defining as discriminatory criteria the fact of being “white” and “men”.
In our experience, in an excess of confidence, we asked ChatGPT, given a link to a generic website, to give us back the main digital barriers; a task that we usually face within usability tests or expert analysis. The result, as expected, was relatively poor, with generic barriers, vague and too distributed among the sites.
These examples demonstrate that in reality this is still an evolving technology and that, paradoxically, what limits it most are errors and prejudices contained within the learning data and therefore: “human”.
In conclusion, this type of neural network is certainly something that is laying the foundations for future digital communication, some say that even search engines will be supplanted, or rather incorporated by these conversational models. However, at the moment it still remains a very advanced chatbot but which has all the credentials to revolutionize the digital world as we know it now; but only time will tell.
Cade Metz., The New York times: “The New Chatbots Could Change the World. Can You Trust Them?”
Josh Bersin., Insights on Corporate Talent, Learning, and HR Technology: “Understanding Chat-GPT, And Why It’s Even Bigger Than You Think.”