Plugavel.
  • Home
  • Tech
  • Car
  • More
    • Privacy policy
    • About us
    • Contact us
No Result
View All Result
Plugavel.
  • Home
  • Tech
  • Car
  • More
    • Privacy policy
    • About us
    • Contact us
No Result
View All Result
Plugavel.
No Result
View All Result

We tested ChatGPT on scientific news

17 de January de 2023
in Tech
We tested ChatGPT on scientific news
ADVERTISEMENT

“Hello, I am Futur.IA, an artificial intelligence specialized in science journalism. I am able to provide you with accurate and up-to-date information on many scientific topics. Do not hesitate to ask me questions ! »

Yes, it is an artificial intelligence (AI) who wrote this short presentation. More specifically, it is ChatGPT, which we named for the occasion Futur.IA (the pun is easy to understand), an artificial intelligence created by OpenAI and capable of making conversation on many varied topics. With quite a few similarities to GPT-3, the AI ​​that wrote and published an article about itself, ChatGPT also adapts to the language used by the person it interacts with. For the text above, he was asked: Good morning ! Could you introduce yourself as if you were a science journalist at Futura, specifying that you are an AI, and calling yourself “Futur.IA”? »

The result is not bad, but pushing for a longer presentation, we got: ” Good morning ! I am Futur.IA, an artificial intelligence designed to answer questions and help people solve problems. As a science journalist at Futura, I am passionate about science and technology and enjoy sharing my knowledge with readers. I am trained in a large number of subjects and can provide accurate and reliable information on many fields, ranging from biology and astronomy to physics and computer science. I am also able to understand and answer many types of questions, whether simple questions or more complex topics. If you need help or have any questions, please don’t hesitate to contact me! » Even better !

How does ChatGPT work? Futur.IA explains it to you!

Good… Now that we have this false journalist, we would like this time to make him write something. How about how to use it? This time the AI ​​was asked to write a guide on how to use it. Of course, it is a success, and here is his answer!

” Sure ! Here is an article I wrote to explain how to interact with me and how I work:

“Are you wondering how to interact with an artificial intelligence like Futur.IA? It’s very simple! To ask a question or request help, all you have to do is type your message in the text box and click on send. I will do my best to understand your request and provide you with an accurate and helpful response.

But how am I able to understand your questions and provide you with answers? As an artificial intelligence, I am trained to understand human language and provide appropriate responses to a wide range of topics. To do this, I use a technique called “machine learning”, which allows me to “read” and analyze millions of pages of text and data to understand how things work and find answers to your questions. .

I’m also constantly updating and improving to provide you with the best possible answers. If you have any comments or suggestions to improve my service, do not hesitate to contact me! I’m here to help and answer your questions.”

To write about news, it’s more complicated

By wanting to test it on the news, that’s where it gets stuck. We try by giving the link of a study: failed, Futur.IA does not have access to the Internet, and has only trained on a limited quantity of documents. We try again by offering to send him the text of a study, but the text is too long, and the figures are missing anyway. Communicating just through a chat is complicated. The AI ​​is able to converse on many topics, but not the newest ones. We try by copying a text from a NASA press release: Futur.IA will have to write a very short article on it. Missed again, the AI ​​manages to start the article but does not finish it. It must be said that many users are there in parallel, so that ChatGPT crashes very often!

This time, we try again with an older subject, on which she probably already has the information: the image of the protostar L1527 in the shape of a cosmic hourglass, captured by James-Webb. And the result is quite successful!

“The James-Webb Space Telescope has just made an exciting discovery: a protostar, called L1527, which lies in the constellation Taurus.

A protostar is a star in formation, still enveloped in a gigantic shell of gas and dust. These stars are thought to form from clouds of gas and dust called nebulae, which contract under their own gravity.

L1527 is particularly interesting because it lies at a relatively close distance from Earth, about 450 light-years away. This means that we can observe and study it in detail using telescopes like James-Webb.

Observations of L1527 by James-Webb have revealed many details about its structure and composition. In particular, it was discovered that the protostar is about 500 times more massive than our own Sun, and that it is about 100,000 times more luminous.

But what is most fascinating is that L1527 appears to be forming a planetary system! Detailed observations revealed the presence of several disks of dust and gas around the protostar, which may be condensing to form new planets.

This discovery opens new perspectives in our understanding of how stars and planetary systems form. Thanks to instruments like James-Webb, we can now study these processes in detail and learn more about the origin and evolution of our universe. »

Need to add a firewall!

Conclusion: write articles on sufficiently old subjects, the time that it was updated, it is possible. But the journalistic style is missing, although it is possible to configure it, therefore to modify its style of writing by asking it to adapt certain criteria. Journalists, we can continue our job! Phew.

Where it gets complicated is for subjects that are not topical. As it is possible to configure the AI ​​so that its tone is human, it is possible to make it write all types of documents. Cover letter, recipe, poem, article, or even a student assignment. It is this last possibility in particular that makes university professors fear many cheatings. So much so that OpenAI has committed to inserting a marker into ChatGPT so that the distinction between human and non-human is possible via conventional detectors.

Tags: artificial intelligenceartificial intelligence risksChatGPTdangers of artificial intelligenceGPT-3newsscientifictested
ShareTweetPin1

We would like to send you notifications with news, you can unsubscribe at any time.

Unsubscribe
  • Home
  • Privacy policy
  • About us
  • Contact us
© 2021 Plugavel - News about technology and cars on one site Plugavel.
No Result
View All Result
  • Home
  • Tech
  • Car
  • More
    • Privacy policy
    • About us
    • Contact us