When ChatGPT is confidently wrong

Подписаться 90 тыс.

Просмотров 8 тыс.

50% 1

You’ve seen articles on the factual mistakes ChatGPT gives people-how it'll confidently mess up school reports or financial interpretation. If you work in tech or content creation, you should be extra skeptical of relying on ChatGPT results, even when your prompts are clear and precise.
Simon Allardice, Creative Director and Principal Author at Pluralsight, asked ChatGPT a well-scoped, specific, verifiable tech question. ChatGPT was impressive. And impressively wrong about the answer.
Read more about Simon’s test here: / confident-wrongness-ch...
Want to learn the basics of ChatGPT and how to apply it to real-world problems? Take our free course here: www.pluralsight.com/courses/c...
Visit us at:
Facebook: / pluralsight
Twitter: / pluralsight
LinkedIn: / plur. .
Instagram: / pluralsight
Blog: www.pluralsight.com/blog

Наука

Опубликовано:

7 мар 2023

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 8

@projectartichoke 6 месяцев назад

And, sadly, a lot of general technical support websites are just copying and pasting text from ChatGPT into their articles and the information is more often then not massively wrong. Some of them even have ChatGPT text apps at the top of the page displaying wrong answers. General consumers of information are much worse off because of it.

@DZX5000 4 месяца назад

omg i asked it to solve a triangle and it was insisting that the answer was correct. it is getting very concerning.

@lutramage6252 7 месяцев назад

I can't understand why this video has no comments (although there was some odd problem when I started to type a comment and the text generated some weird entries I had not written). This video very clearly and efficiently explains what is a hugely significant issue with using generative AI for technical research - currently the systems are simply not capable of providing robust, authentic, accurate and wholly reliable technical answers to even very straightforward questions, and yet their answers are deceptive and confident even to the extent of providing false (entirely made up) sources and secondary statements that provides an air of authenticity that makes them difficult to distrust. This video should have millions of views, not 4k (as of December 2023).

@pluralsight 5 месяцев назад

Thank you for sharing!