Thanks Dylan, great video! Took your advise and just switched our backend to. Looking forward to the reduced SQL injection. Glad you took the time to delve into this.
Communication isn't just one of the most powerful skills you can cultivate as a developer, but *as a person*. And in the particular context, it seems to me that being able to communicate well in the submission not only gets the idea across, but reassures the program committe you'll probably be able to communicate well in the *talk*.
What a great analysis! I did not expect this problem with auto-generated texts would be that big. And yes, communication is a must, but it's very difficult to grow in yourself. I'm trying hard, but still not as good as I want! 😄
Interesting insights, I really like how you delved into the landscape around chatGPT's generative AI being used to enhance peoples poor writing skills when creating abstracts, this really empowered me. :P (I tried to sneak "navigate" in there but I couldn't without making it sound wrong somehow)
The CS-ignorant use of the word "capability" is a fun tell for chatgpt, too. There have been some great talks on doing capabilities in python, especially around CapnProto (and hopefully soon, WASM), but seeing the term used as a synonym for "feature" in a software abstract is pretty jarring.
So... perhaps this will grow a need for a "web of trust". Had to google it, because it was long ago, but had to do with pgp i believe and ways to tell who is who, with people basically being able to filter out trusted people. I mean, won't it be nececery to eventually lean on such tech to know that the article we read actually has some quality based on who wrote it as opposed from the bad stuff that probably will flood the net?
I think the only sensible use for these nonsense generators is to generate something, then actually read the output and realise this sounds awful, and then think "I can write this much better myself"....
Why does everyone always pick Perl for this? You can do stupid things in any language. Every language is equally vulnerable to SQL injection attacks, because that's a misuse of string construction, which every language can do. Perl+CGI is, presumably, a proxy phrase for things like Matt's Script Archive, but it isn't inherently a bad idea to use either of those things, any more than it's a bad idea to use PHP, which has a far worse reputation and yet it's still all over the place.
Because when we learned Perl, we learned what to do and what not to do in, "Learning Perl" and other classic textbooks of the time. It is because the author of Perl wrote about these problems, and our understanding of those problems are often based on Perl examples. You are 100% correct, of course.
Ok, a variation of "diggy diggy hole" is just about the last thing I expected this video to start with...
The Yogscast truly has become a foundational pillar of Internet culture.
I bet Dylan heard it from Wind Rose first
Thanks Dylan, great video! Took your advise and just switched our backend to. Looking forward to the reduced SQL injection. Glad you took the time to delve into this.
Communication isn't just one of the most powerful skills you can cultivate as a developer, but *as a person*. And in the particular context, it seems to me that being able to communicate well in the submission not only gets the idea across, but reassures the program committe you'll probably be able to communicate well in the *talk*.
Delving into delving! Thank you!
What a great analysis! I did not expect this problem with auto-generated texts would be that big. And yes, communication is a must, but it's very difficult to grow in yourself. I'm trying hard, but still not as good as I want! 😄
Interesting insights, I really like how you delved into the landscape around chatGPT's generative AI being used to enhance peoples poor writing skills when creating abstracts, this really empowered me.
:P (I tried to sneak "navigate" in there but I couldn't without making it sound wrong somehow)
I am so fucking glad I was smelling the ChatGPT reveal from the beggining of the video and when I saw delve oct 2022 I was like, that's it.
The CS-ignorant use of the word "capability" is a fun tell for chatgpt, too. There have been some great talks on doing capabilities in python, especially around CapnProto (and hopefully soon, WASM), but seeing the term used as a synonym for "feature" in a software abstract is pretty jarring.
This talk had a bigger twist than an M Night Shyamalan film.
This here, this is 'Java'. And my Spidey senses were tingling, anticipating the next Java joke. ^^
So... perhaps this will grow a need for a "web of trust". Had to google it, because it was long ago, but had to do with pgp i believe and ways to tell who is who, with people basically being able to filter out trusted people. I mean, won't it be nececery to eventually lean on such tech to know that the article we read actually has some quality based on who wrote it as opposed from the bad stuff that probably will flood the net?
I should talk about webs of trust in a future video - I actually went to a PGP key signing party once!
That’s unfortunate
I think the only sensible use for these nonsense generators is to generate something, then actually read the output and realise this sounds awful, and then think "I can write this much better myself"....
TFIDFs not dead 😂
Why does everyone always pick Perl for this? You can do stupid things in any language. Every language is equally vulnerable to SQL injection attacks, because that's a misuse of string construction, which every language can do.
Perl+CGI is, presumably, a proxy phrase for things like Matt's Script Archive, but it isn't inherently a bad idea to use either of those things, any more than it's a bad idea to use PHP, which has a far worse reputation and yet it's still all over the place.
Because when we learned Perl, we learned what to do and what not to do in, "Learning Perl" and other classic textbooks of the time. It is because the author of Perl wrote about these problems, and our understanding of those problems are often based on Perl examples. You are 100% correct, of course.