Testing ChatGPT!

What should I ask this AI?

This is ChatGPT, which was released this week by OpenAI and is based on their language model GPT-3. It allows you to chat with an AI. We're still learning about what it can do, so let's ask it questions and see how it responds.

I made a longer episode about OpenAI's image generator, DALL-E 2, which you can watch here:

#shorts #ChatGPT #AI #artificialintelligence #openAI

[beng]

I'm a teacher and I just tried this with an essay question I ask my students. The response was brilliant - undetectable even! Then I asked it to grade its own answer and provide feedback which it did flawlessly. So I asked it to write a report on a student which it also did with flying colours. To say this will impact on education is an understatement. It will redefine it!

[Bryan-fbdh]

This is like a calculator for life. Literally. What the?!

[mosesgaming]

Dang.... Its crazy to look back and see just how cool it was when it first came out and its still so cool

[fivetimesyo]

Thanks, I'm gonna go collect my PhD now.

[saimanish]

I told the AI to write a 1000 word essay on "Harmful effect of AI on humankind" 😂😂😂 and it did write it

[digiscream]

My wife trains people to get health and social care qualifications, and part of that means using plagiarism scanners. We gave it four questions, with five-paragraph answers, and...it scored 0% plagiarised on every single one.

Insanely impressive technically, and utterly terrifying for the education establishment.

[rileysimpson]

The “Robot that does your homework” is actually here

[Marshmellow]

I asked it to write ten lines of dialogue between a customer and a cashier at a supermarket and chose two topics of conversation: The price of avocados, and the weather in Wisconsin. When I asked it to use both, it gave me 13 lines of dialogue. When I gave it the choice between either, it chose avocados and was able to complete the scene in 10. I asked it to explain its choice, and it said it couldn’t find a way to make the dialogue sound natural in ten lines using both topics, but when given the choice the avocado topic could be used go create ten natural sounding lines of dialogue. The implications of these responses are incredibly fascinating to me: we learn A. It can make decisions B. It can choose to ignore certain parameters of your request C. Why it chooses to ignore those parameters and D. That it can explain its own thought process. It doesn’t quite pass the Turing test, but it’s scarily close imo.

[renganomics]

getting this in my feed now after all the essay scandals and whatnot is pretty cute to see all the early excitement

[JesusIsTheGOAT]

I can't imagine how easy high school and college would have been for me if this had been formed at least 8-10 years earlier.

[ishaankotecha]

I just typed in questions about my dissertation and it starting telling me what papers to read this is nuts.

[michmach]

Sweet nostalgia, I remember the first time I discovered ChatGPT. It was still pretty early on, but I was so hyped about it. Thanks for the memories, Cleo. I know it's only been 2 years ish but still. Feels like a lot has changed.

[Owen_loves_Butters]

You know what I mostly use it for? Writing stories. When you want to read a story with a specific plot, that's pretty much impossible unless you know someone that writes or are incredibly lucky. But with ChatGPT it's really easy, and even if you know someone that writes, it still writes it in seconds instead of hours.

[ajbp]

This is insane! Was impressed by both replies, but on different levels.

[TheZilacshack]

I think that most teachers will be able to catch an A.I written essay, when their student’s writing suddenly becomes immensely more intelectual and professional.

I don’t think that’s actually a smart Way to use it, as a student, instead of plunging through a million websites (most of which are banned on school laptops), it’s way easier to read a few paragraphs that have been neatly compiled by A.I, and put it into your own words! It really makes homework so much faster, and I really don’t feel like I am( or can be blamed) for plagiarising, as I simply retrieved the knowledge I was meant to retrieve anyway!

[ryanflynn]

Exact reason they need to change the curriculum. Pointlessly memorizing something just to forget it has always been silly and now it’s more useless than ever. They need to teach kids to think critically

[dAvrilthebear]

What bothers me is that this AI exposes how often humans actually simulate intelligence. Writing school essays is a good example. We make our kids pretend for 5 paragraphs that they sound like adults experienced in literary criticism and in life in general. But in fact when you are 15, you have only a very limited understanding of what it's like to be someone like Ghatsby. This is why school essays are mostly generic, unoriginal and insincere -- they are designed to be this way. This is why an A.I. excels in such a task.

[amscuda]

This is literally the future, we’re finally here

[JJs_playground]

This is going to have very negative implications for school. Even if the essay is not perfect, it's good enough to get the student started and just fill in a couple of key points in their style to pass it along as their work. With chatGPT and DALLE we are entering a very interesting world that I thought was going to be in the far, far future, but it's here now.

[Aldraz]

Schools are done. Finished. Teachers will lose their jobs. GG ez.