How did it come to this?

OpenAI is an artificial intelligence (AI) laboratory in San Francisco, co-founded in 2015 by Elon Musk and the entrepreneur programmer Sam Altman, eager to create the first AI as intelligent as humans. Then funded to the tune of 1 billion dollars by Microsoft, OpenAI has just received an investment offer of ten times this amount from the computer giant which wants to boost its Office suite with its tools. A growth that is phenomenal to say the least!

OpenAI, which is also behind the surprising Dall-E image generator, trains AI technologies from digital data. In 2015, the research organization developed a text generator, GPT, Generative Pre-trained Transformer . It initially involved training the prediction of the next word in a sentence, using huge quantities of digital texts (books, web articles, blogs). The system then swallowed more than 40 GB of data from 8 million pages.

But it has become so adept at mimicking the way the human brain gathers information to write an analysis, a speech or even a poem that research has been suspended until the full extent of its reach and possible ethical implications can be determined.

Then GPT-3 was born, the largest language model with 175 billion parameters and relentless precision. ChatGPT is the version available to the public since November 30. In less than a week, it reached one million users, which has continued to increase since then. A paid professional formula is about to be implemented.

What is ChatGPT?

This conversational interface allows you to answer questions and engage in dialogue: ChatGPT has ideas, can tell jokes, come back to points raised and even admit its mistake! It's your new friend.

ChatGPT synthesizes and composes text at a phenomenal speed. Just enter simple instructions to receive a coherent piece of writing, in the desired style. Summarize the highlights of a report, do an assignment or compose a song, the possibilities are endless. Its creativity is impressive. And the result is amazing!

Much to the surprise of those who created it, it can even generate computer codes, write programs and solve their puzzles.

By using advanced machine learning techniques, ChatGPT is able to constantly adapt and improve itself. It’s powerful.

It is becoming possible to create more "real" and more efficient virtual assistants. Could customer service, marketing, journalism, artistic content creation or even teaching be threatened?

ChatGPT, too efficient?

It's mind-blowing to see a text written with impeccable grammar, synthesizing a quantity of information in a few seconds. Enough to make the most seasoned writers shudder... and even Google! However, ChatGPT does not redirect to a list of sites: it directly gives the answers to the question asked. Is this the end of the supremacy of the most famous search engine?

And what does the education community think about it? It's all hands on deck! In New York schools, ChatGPT has already been banned. While students are delighted with this super text-generating chatbox , it seems a priori to threaten the practice of teaching, at least in its current form.

Indeed, plagiarism is not easy to thwart, but on closer inspection, the answers are not so perfect. OpenAI has even acknowledged that if the form is beautiful, the content is not always: the sentences are well-crafted, but the facts can be wrong and the point, absurd. The application also does not hesitate to create new sources or cite characters or authors who do not exist. The level of language used can be a clue for the teacher: same grammatical constructions, same types of personal examples, reasoning carried out in the same order, with the same qualities and the same defects...in short, a new kind of cheating.

It seems that there is a great opportunity for the teaching community to review certain practices, including the assessment of skills, and to put critical thinking and problem solving, so useful in life, back on the agenda!

The same goes for the information, marketing and creative industries. Artists, writers, composers and other creators are concerned that AI will use their work without their consent and without providing them with compensation. And rightly so.

As for journalists, AI systems can provide research and analysis, but GPT-3 is not capable of reasoning and has no representation of the world. It cannot tell the difference between what is true and what is false! The information seems correct, but there is no fact-checking per se. AI can also reflect societal, cultural, racist or sexist biases. Disinformation is possible; beware of the risk of massive propaganda. Finally, the data that ChatGPT uses dates from 2019, before the pandemic!

It remains that the tool could play a useful role in journalistic routines of research, alerts, real-time information overview or even basic text to enrich, but the effectiveness will depend on the quality and availability of the data.

With their emotional and sensory intelligence, their critical mind, their memorization of memories, complex social and spatial interactions, and their level of consciousness, humans naturally stand out in front of AI and its immense power of speed of execution and data storage. If AI is a source of technological development with incredible possibilities, some fear opening a Pandora's box... It is certainly hand in hand that the two will have to clear the way for the future.