I want to sign up for the beta but I am put off by the strict limitations of approved apps in their use case guidelines. This technology is so powerful so I understand why they might want to limit certain areas, but to draw limitations and cut off millions of potential developers/users because their use case could possibly not be human monitored is interesting. At least they could reject certain requests in the backend rather than cut off an entire audience of users due to a supposed guideline. But Elon Musk is one of the founders, and his ego definitely wants to make Open AI his own formulated mission. GPT-3 (or something very similar) needs to be fully open sourced.
Yeah the restrictions take a bit of the fun out of it, but its still fun to play with even if you're just using the built in playground environment. I hope to see a comparable project open sourced as well!
Yeah, let us include "Uber-Clippy", the endboss into our product. What could possibly go wrong? Jokes beside, thank you for the video and the in-depth explanation, CttM:)
Thanks for sharing. GPT-3 is really crazy. Almost as cool as the image gen stuff. Does this allow the opposite - interpreting text or speech to determine if someone likely answered a question correctly (I would see that as huge in gaming). Like, you configure - we're looking for certain words in answer. If they get at least "x" of those, then they answered correctly. I'm thinking in terms of VR where player makes some decisions or actions in game based on speech - regardless of whether the speech to text is a local transformation or part of the service.
I agree, it's pretty impressive! I think GPT-3 is quite capable of what you're saying - you can give it instructions on how to answer in plain language. For example your input could simply be "The question is x. the user's answer was y. The real answer is z. Did they answer the question correctly?" and GPT-3 should spit out a simple "yes" or "no" response. It might require a bit of tuning in terms of how close the user's answer needs to be to be considered correct, but it's definitely something GPT-3 is capable of.
@@codetothemoon - Thanks, I was thinking about it later, and determined that all I would really need is speech to text capabilities, then interpret the text for context and words, but GPT-3 would still be very useful for tons of stuff - really haven't thought of all the possibilities, but I know people will come up with amazing games using gpt-3. Rust has a wrapper for "deepspeech" library, deepspeech-rs, which is perfect.
Thanks! I actually make all of my thumbnails with Adobe Illustrator, and I source many of the graphics from flaticon.com. In this particular one, the chip with the brain on it is from flaticon.com.
Im a very green rustacean, and only dabbled in Conversational ai via rasa about 3 years back... My follow up uestion to this video is how could one implement fully localized version of this chatbot? Circumventing the need to be online
At first glance I thought it's functionality also existed in the "Rust" extension, but I didn't investigate too deeply. What set of VSCode extensions would you recommend for Rust development?
Great question - i haven't actually read the whitepaper, but my guess is that the concept of "attractiveness" and examples of humor both exist in the training data. I believe some flavor of deep neural network is trained on this data, so it can then infer appropriate responses to a prompt based on what it saw in the training data (even if that specific prompt didn't exist in the training data).
Yep the responses are generated by gpt3 based on the users input and a static hardcoded string appended to that input. I said the responses weren’t pre-generated, not that they weren’t influenced by the input (all AI is and must be)
I copy and paste the rustopenai code in my linux terminal, I run it , and I got this message !! thread main packinked at src/main.rs:42:51 called Result::unwrap() on an `Err` value: NotPresent note: run with `RUST_BACKTRACE=1` enviroment varialble to display a backtrace can you help please
keep these tutorials going, really helpful and enjoyable to watch!
Thanks will do, glad you found it useful and entertaining!
This video is outstanding !
Easy to understand, super interesting and fun to watch !
Thank you, your description is precisely the goal I was shooting for when I made it. Glad you found it valuable!
😂😂😂 LOL, that was hilarious. OpenAI loves Rust now
Thanks Anotidaishe, glad you liked it! Wasn't sure if folks would appreciate the humor ;)
Wonderful. Very instructive on a whole lot of levels. Thank you!
Thanks, glad you found it valuable!
This channel is insane! Thank you for what you are doing!!!
Thanks for watching Jack! more to come!
Regarding printing a newline, you can omit the quotes: `println!()` will print an empty line.
ahh nice thanks for the tip!!
Loved it, Haha. Got me a smile.
Nice, that's what I was going for!
I want to sign up for the beta but I am put off by the strict limitations of approved apps in their use case guidelines. This technology is so powerful so I understand why they might want to limit certain areas, but to draw limitations and cut off millions of potential developers/users because their use case could possibly not be human monitored is interesting. At least they could reject certain requests in the backend rather than cut off an entire audience of users due to a supposed guideline.
But Elon Musk is one of the founders, and his ego definitely wants to make Open AI his own formulated mission.
GPT-3 (or something very similar) needs to be fully open sourced.
Yeah the restrictions take a bit of the fun out of it, but its still fun to play with even if you're just using the built in playground environment. I hope to see a comparable project open sourced as well!
Wonderful 😌
Great vid, ty! 🙋🏻♂️
Thank you and thanks for watching!
Yeah, let us include "Uber-Clippy", the endboss into our product. What could possibly go wrong?
Jokes beside, thank you for the video and the in-depth explanation, CttM:)
Wow ! Thank you for example AI integrate with Rust ! Hey !
Glad you enjoyed it, thanks for watching!
Hahahahah so funny. Love this channel and love the rust content!!
🙃 Thanks Eran!!
Thanks for sharing. GPT-3 is really crazy. Almost as cool as the image gen stuff. Does this allow the opposite - interpreting text or speech to determine if someone likely answered a question correctly (I would see that as huge in gaming). Like, you configure - we're looking for certain words in answer. If they get at least "x" of those, then they answered correctly. I'm thinking in terms of VR where player makes some decisions or actions in game based on speech - regardless of whether the speech to text is a local transformation or part of the service.
I agree, it's pretty impressive! I think GPT-3 is quite capable of what you're saying - you can give it instructions on how to answer in plain language. For example your input could simply be "The question is x. the user's answer was y. The real answer is z. Did they answer the question correctly?" and GPT-3 should spit out a simple "yes" or "no" response. It might require a bit of tuning in terms of how close the user's answer needs to be to be considered correct, but it's definitely something GPT-3 is capable of.
@@codetothemoon - Thanks, I was thinking about it later, and determined that all I would really need is speech to text capabilities, then interpret the text for context and words, but GPT-3 would still be very useful for tons of stuff - really haven't thought of all the possibilities, but I know people will come up with amazing games using gpt-3.
Rust has a wrapper for "deepspeech" library, deepspeech-rs, which is perfect.
Super duper awesome ❤️
Thanks, glad you liked it!
where did you make your thumbnail? really cool!! 🔥🔥🔥
Thanks! I actually make all of my thumbnails with Adobe Illustrator, and I source many of the graphics from flaticon.com. In this particular one, the chip with the brain on it is from flaticon.com.
Lovely!
Thanks Karl!
Wow. Thank you
Thanks for watching!
Which text to speech program did you use?
Amazon Polly
Thanks for the great tutorial. Any chance we could get the GH repo for this? I'm getting a bunch of errors and would like to match my code with yours.
Thanks! The code is available here github.com/Me163/youtube/tree/main/RustOpenAI
Im a very green rustacean, and only dabbled in Conversational ai via rasa about 3 years back... My follow up uestion to this video is how could one implement fully localized version of this chatbot? Circumventing the need to be online
Why are you not using `rust-analyzer`?
At first glance I thought it's functionality also existed in the "Rust" extension, but I didn't investigate too deeply. What set of VSCode extensions would you recommend for Rust development?
It a good boy for advertising rest
Thanks for watching, I strive to be a good boy
awsom
Thanks for watching!
is this on github anywhere?
How does GPT-3 understand what humor is?
Great question - i haven't actually read the whitepaper, but my guess is that the concept of "attractiveness" and examples of humor both exist in the training data. I believe some flavor of deep neural network is trained on this data, so it can then infer appropriate responses to a prompt based on what it saw in the training data (even if that specific prompt didn't exist in the training data).
@@codetothemoon Wow! Thanks for your answer 👍.
Completely generated by the A.I without influence,.. oh by the way I hardcoded it to be biased.
Yep the responses are generated by gpt3 based on the users input and a static hardcoded string appended to that input. I said the responses weren’t pre-generated, not that they weren’t influenced by the input (all AI is and must be)
@@codetothemoon
Not all A.I. relies on input, just wait for July 6th.
Couldn't you have just asked OpenAI to write all that code for you? lol
lol possibly! This video was made in the GPT-3 era, and I wasn't fully aware of what it was capable of...
I copy and paste the rustopenai code in my linux terminal, I run it , and I got this message !!
thread main packinked at src/main.rs:42:51
called Result::unwrap() on an `Err` value: NotPresent
note: run with `RUST_BACKTRACE=1` enviroment varialble to display a backtrace
can you help please