ChatGPT : Boarding for Better and for Worse!

ChatGPT, l'IA qui pourrait changer le monde

ChatGPT, the “chatbot”, or conversational agent developed by Open AI has become “THE” star of social networks. His skills and his “finds” reveal a tool of formidable efficiency. You can ask ChatGPT for almost anything: writing a poem, a speech, an article, a song, in almost any language. It can also help write codes and programs that can be used by cybercriminals and online scammers to generate scam and phishing emails and thus increase threats such as ransomware. It must be said ChatGPT is not the first conversational agent, far from it. Because other examples of artificial intelligence in other areas are already operational.

 A real journey for better and for worse.

Let's start at the beginning, what is ChatGPT?

ChatGPT is a language processing model developed by OpenAI, a California start-up. It is trained to understand and generate text using a large amount of training data. It can be used for tasks such as content generation, code generation, question answering, machine translation and much more. It uses a machine learning technique called “style transfer” that allows it to adapt to different styles of language and context

Research: ChatGPT can be used for research in areas such as natural language understanding, AI, and machine learning.

Computer developers and coders use Chat GPT to automatically generate code for computer algorithms and programs. Although he himself is a “chatbot”, he is able to create intelligent “chatbots” for various uses.

Some application examples of ChatGPT

This conversational agent is undergoing constant development and its applications are therefore likely to evolve and be enriched.

At this stage, it can for example generate various content ranging from dialogues, scripts, stories, articles, texts. Once the user enters a few keywords, the chatbot almost immediately provides content “proposals”.

It also allows you to translate any content from one language to another. We did a test on four languages, and we obtained the translation of a text in French into English, German, and Spanish. To our question: “Into how many languages ​​can chatGPT translate a given text?” “. The response was: “ChatGPT is a language processing model designed to understand and generate English text. It is not specifically designed for language translation. However, it can be used in combination with other translation tools to help translate text into other languages.

Thanks to its learning capabilities and “ingested” data, it also responds instantly to questions of all kinds. To the question: “Why should I use chatGPT?” the answer was:

“There are many reasons why you might want to use ChatGPT:

Text generation: ChatGPT can generate text from given prompts, which can be useful for writing content, generating scripts or creating auto-responses.

Natural language understanding: ChatGPT is able to understand natural language, which means it can help analyze and process textual data, such as sentiment analysis.

Entertainment applications: ChatGPT can be used to create text-based stories, jokes and games to entertain users.

Improved productivity: ChatGPT can be used to automate certain tasks, such as writing emails or preparing reports, which saves time and improves productivity.

ChatGPT is a boon for cybercriminals.

This chatbot is a real boon for cyber criminals as it has started being used for evil purposes. Indeed, if the OpenAI tool is mainly used for peaceful purposes, ChatGPT can also be used to develop malware and other malicious applications such as:

– Generation of phishing emails to steal personal information or login information.

– Generation of automated messages to send spam on a large scale.

– Generation of false content and false news or opinions to manipulate public opinion or harm a company or person.

– Creation of malware and malicious code to infect computers or networks.

Faced with the success of the conversational agent, threat actors are interested in it. Check Point Research researchers Check Point Research have found three cases where black hats leveraged ChatGPT for malicious purposes.

What data is ChatGPT trained on?

The “chatbot” is an automatic language processing model based on neural networks, which is trained using large amounts of training data. These data come from natural language texts, newspaper and magazine articles, books, web pages, discussion forums.

This data is either public accessible without any restrictions, or proprietary data (using tools such as web crawlers or scrapers to collect web pages or social network data), or proprietary data (chat data from a service customer or company email data).

It is important to note that there is a risk that ChatGPT will produce duplicate data, hence the importance of carrying out some checks before using the data provided by the “chatbot”. For this, the training data must be checked using duplicate content and plagiarism detection tools.

It is equally important to check the results output from ChatGPT in order to detect errors or potential problems.

Finally, it is important to note that, faced with the use of the “chatbot” for the purposes of exams and school and university work, institutions are beginning to react, such as SciencesPo Paris, and schools in New York.


According to STATISTA of January 30, 2023 (see chart hereafter), ChatGPT would have attracted one million users in five days, while it took 2 years for Twitter to reach this level.

 

Selon STATISTA du 30 janvier 2023, ChatGPT aurait attiré un million d’utilisateurs en cinq jours, alors qu’il a fallu 2 ans pour Twitter pour atteindre ce niveau.