Written by Lucas Décorne,
Reviewed by Arthur Puybertier,
Translated by Lucas Décorne.
​
Published on January, 28th, 2023.
ChatGPT was released in November 2022 by the OpenAI artificial intelligence lab from San Francisco, CA. The new version, GPT-3, marks a new era of AI.
​
An outstanding invention
​
On November 30, 2022, the general public discovered the power of ChatGPT, a freely available tool that generates answers to short prompts. The chatbot is able to write songs, motivation letters, or even software programs with disconcerting ease. It can restore the context of many texts and summarize them in the same way as humans.
​
The intelligence scoured the internet, analyzing thousands of digital books, the entirety of Wikipedia, and nearly a trillion words posted on social media and blogs. This system hides many years of work inside the world’s leading AI labs, including OpenAI, which was funded by over $1 billion dollars from Microsoft. In December, ChatGPT had already been surveyed by no less than one million users.
​
AI forces a shift in the way we educate
​
In the United States, as soon as the chatbot came out, professors began to think about a new way of teaching. They have to restructure their courses and take preventive measures to counter the use of AI to adapt to this new era in sight. While the chatbot was only released a few months ago, some professors noticed some exceptional papers in their classes. These were not written by their students, who actually used ChatGPT to write their papers. Teachers will need to move toward more oral exams, group work, and handwritten assessments instead of typed ones and rather phase out take-home, open-book assignments.
​
If ChatGPT can be used as a time-saving or brainstorming tool in an effective way, it will never create work equal to that of an expert, especially for generating complex programs. It makes mistakes by sometimes incorrectly explaining ideas and misquoting sources because information on the internet is not always reliable and technology is not capable of discerning truth from falsehood.
​
Another problem arises at the ethical and legal level
​
The content provided by ChatGPT could infringe on intellectual property rights by being strongly inspired by existing written content or images. Legal uncertainty also remains concerning the protection of personal data, potentially recoverable by OpenAI. A new regulation on AI systems should be adopted in 2023. The objective would be to ensure that AI systems placed on the EU market are subject to strict laws to limit their use because without legal basis and without ethical design it is bound to lead to derives.
While GPT-3 was released in November, OpenAI is expected to release the GPT-4 version soon, which is better at generating text than previous systems.
​
​
​
Lucas Décorne