Artificial Intelligence General Discussion

[PluMGMK]

Oh that's right yeah, one of the guys I work with used it to generate a suitably-dystopian image to accompany a certain phrase that had been trotted out (It's innocuous enough but I don't want to post it here in case somebody I know IRL somehow reads this )

[Elite Piranha]

I found this late, but I think it was worth sharing here:

Bad news for execs in Hollywood that want to replace writers with AI. On Friday, a judge ruled that a piece of art created by AI is not open for copyright protection. “Human authorship is a bedrock requirement” when it comes to copyrighting art. This ruling is huge for writers and actors currently on strike. If AI work is not protected, it could raise issues for studios.

Source: https://www.facebook.com/photo/?fbid=68 ... 6946311108

Comedian Sarah Silverman and authors Richard Kadrey and Christopher Golden have filed two copyright infringement lawsuits, one against Facebook parent Meta Platforms and one against ChatGPT-maker OpenAI for using their work to train chatbots.

Source: https://www.pcmag.com/news/sarah-silver ... fringement

[SilverLum]

I found this late, but I think it was worth sharing here

This is great news! I’m glad that the rights of real writers are being protected, and it’s a good thing these laws against AI are being passing quickly.

[PluMGMK]

So basically anything that's created by an AI is just a derivative work whose copyright belongs to the author of whatever it's based on? That's definitely gonna make life complicated for people who try to use AI to make stuff!

Also, I heard about this and meant to find and post it during Record Day:

[DaveRattlehead]

That will definitely keep the work for the talented people. Unfortunately there are "wonderful" people who would substitute a professional writer because "an AI would do it better"

That Barbie Girl cover reminded me of this piece of art:

[Elite Piranha]

I've been seeing a lot of AI art in DeviantArt recently, and in the profile of one of the accounts responsible for creating these images, I found the text: "PLEASE DO NOT REPOST MY IMAGES". I don't know how to feel about that, can you really say that they're your images?

I'm so confused right now .

[DaveRattlehead]

AI gets trained with that art, so there's a small possibility to get caught by the original author (or reference, however you'd prefer to call it).

AI is growing by leaps and bounds. Relatively recently it was released to the public an AI which is able to translate videos into other languages in a completely natural way. To be honest, there are a lot of funny videos about it, but this leads me to the next question... Will some smartass come along and try to replace dubbing actors with this?

[Elite Piranha]

Well, I heard that AI still can't entirely replace voice actors due to some shortcomings related to inflection, tone, and naturalness. However, like in other areas, it can decrease the number of people that are needed to complete a project (for example instead of hiring 10 people, you only hire 2 because AI can do the rest of the job). This could be beneficial for smalls teams with very limited budget, but it could also incentivize companies to try to reduce the number of human workers as much as possible to lower costs.

Bump: I was thinking, in the past artists used watermarks to protect their works and hid full-resolution images behind paywalls, but now you can use AI to remove watermarks and upscale images.

Also, it seems that: "Even Microsoft Paint is getting an AI art generator"

Microsoft says it is currently testing a preview of the tool, which again will be powered by DALL-E. Named Paint Cocreator, the feature will allow users to generate images based on a text prompt or select from a range of art styles. Paint Cocreator will generate three variations, and users will be able to choose the one they like and then edit the artwork by adding layers – another novelty for Paint.

[DaveRattlehead]

Well, it was a matter of time, but we finally saw the first AI scam involving a famous youtuber

[Elite Piranha]

I remember something similar happening earlier this year, about a fake ad of Joe Rogan endorsing Alpha Grind:
https://mashable.com/article/joe-rogan- ... eepfake-ad

[DaveRattlehead]

This is not revolutionary news, but I would like to share this with you, I think it might be interesting.

A few weeks ago I mentioned that I was going to teach some lab classes at the university. As homework, the students had to make a report and send it to me in the virtual classroom for me to correct their work. Well, I have detected a report that is clearly made with ChatGPT

I don't have any tools to show it was ChatGPT generated text, but it's very, very clear. I don't want to be the bad guy, but I don't think this is right... What would you do in my situation?

[Elite Piranha]

I guess you could use tools like TraceGPT to further back up your claims:
https://plagiarismcheck.org/ai-plagiari ... -detector/

[DaveRattlehead]

Well that was very useful, thank you! My suspicions were right and their text was 99.83% potentially AI generated. I also did a test with my template and the TraceGPT didn't detect it as AI generated. I have to think what to do, I didn't expect to see something like that at the very first

I also had another proof before using that tool... It seems ChatGPT can't do an exercise properly, and when I tested it yesterday it showed me the same mistake they did

[PluMGMK]

So… Are you gonna give 'em a zero?

[DaveRattlehead]

To be honest, I don't know what to do in these cases. As I said, I wouldn't have expected to see that in my very first time doing this . At the moment, I simply didn't validate all of their AI generated answers, but I guess I should discuss what to do with the coordinator.

After this, I wonder how many times they have managed to cheat until I caught them

Edit (17/10/23): Well, there are some news about this... The coordinators don't know what to do as they can't detect the people using the AI as they're not used to it. I was told not to give them a 0, but to give them a warning. In the end, I didn't validate the AI generated answers, that was my warning .

Fun fact after all that madness: this was unintentional, but... If I plot the students' grades, I get a Gaussian shape as a result

[PluMGMK]

So what do people make of the bizarre situation unfolding at OpenAI? https://news.yahoo.com/twists-turns-ope ... 33358.html I can't say I've really wrapped my head around it, but I just wanted to bring it up…

We were talking about it at lunch today, and for the first time somewhat-realistic scenarios for a "Singularity" started to form in my mind. I'm sure it's nonsense, but…

[Elite Piranha]

Yeah I heard about that story, some quotes from these articles really make me feel uneasy:

Ahead of OpenAI CEO Sam Altman’s four days in exile, several staff researchers wrote a letter to the board of directors warning of a powerful artificial intelligence discovery that they said could threaten humanity, two people familiar with the matter told Reuters.

Some at OpenAI believe Q* (pronounced Q-Star) could be a breakthrough in the startup's search for what's known as artificial general intelligence (AGI), one of the people told Reuters. OpenAI defines AGI as autonomous systems that surpass humans in most economically valuable tasks.

Source: https://www.reuters.com/technology/sam- ... 023-11-22/

The same article says that the capabilities of Q* claimed by the researchers, could not be indepently verified though.

[Elite Piranha]

Something that I heard about recently:

Apple made an AI image tool that lets you make edits by describing them

Apple researchers released a new model that lets users describe in plain language what they want to change in a photo without ever touching photo editing software.

The MGIE model, which Apple worked on with the University of California, Santa Barbara, can crop, resize, flip, and add filters to images all through text prompts.

[Hunchman801]

It also surprised me it has integrated DALL-E technology, although it still doesn't seem to work very well... Here's a photo of Rayman smoking

I believe this happens because copyrighted characters such as Rayman are excluded from most training sets to avoid lawsuits. As a consequence, the model can only think of Rayman as a generic cartoony character.

I guess you could use tools like TraceGPT to further back up your claims:
https://plagiarismcheck.org/ai-plagiari ... -detector/

This seems to work really well, I have tested it on both human-written and ChatGPT-written text, as well as human-edited versions of the latter, and I can totally see a pattern in the spectrum of authenticity scores provided by the tool.

Apple made an AI image tool that lets you make edits by describing them

Interesting! This should integrate well with the phone's assistant.

[DaveRattlehead]

It also surprised me it has integrated DALL-E technology, although it still doesn't seem to work very well... Here's a photo of Rayman smoking

I believe this happens because copyrighted characters such as Rayman are excluded from most training sets to avoid lawsuits. As a consequence, the model can only think of Rayman as a generic cartoony character.

That explains why you could generate Pixar-style images, but not specific Pixar characters

I guess you could use tools like TraceGPT to further back up your claims:
https://plagiarismcheck.org/ai-plagiari ... -detector/

This seems to work really well, I have tested it on both human-written and ChatGPT-written text, as well as human-edited versions of the latter, and I can totally see a pattern in the spectrum of authenticity scores provided by the tool.

Indeed. I used it a few months ago to detect GPT written text and it was very accurate. As it's a very recent topic, there were no specific rules against AI used for homework, so I couldn't make themn fail the laboratory part of the subject for that. I know there were some updates planned for next year, so I guess there's a very similar situation everywhere.