Artificial Intelligence General Discussion

What are your feelings towards A.I.?

  • Mostly positive

    Votes: 5 13.2%

  • Mostly negative

    Votes: 13 34.2%

  • Mixed/Ambivalent

    Votes: 20 52.6%
  • Total voters
    38
Elite Piranha

Elite Piranha

Messages
3,521
Tings
151,050
A topic to discuss anything related to AI. I'll start by sharing this video:


In an ideal scenario, sites should not be able to use your data/work/art to train an AI without your consent. Even if you have an unique style, if an AI comes an learns to perfectly imitate what you do, it basically renders you useless. Maybe in the future, every artist will have their own AI, trained based on their own work to mimic their style (however I don't know what would be the minimum dataset to make it work) and any revenue produced by the AI will go to the artist (the owner of the dataset).

It also makes me wonder if the data from RPC has already been used to train an AI.
 
Wow… I think Humch has been right calling the site "Deviant Act" for all these years :hap:

GitHub (i.e. Micro$oft) have started doing something very similar earlier this year: https://sfconservancy.org/GiveUpGitHub/
It's led to a push to give up GitHub for good, but I'd say there's too much inertia for many people to actually do that. (I haven't taken any action for example :oops2:) I wonder if dA will be the same…
 
This AI stuff is a bit scary to an extent compared to back when it was literally just a CPU controlling an NPC. Will there actually come a day where you don't get to talk to a human at all on the other side of the line anymore when you ring tech support. :shock:
 
It reminds me of Shoshana Zuboff's book on surveillance capitalism I read a few years ago. She used the term "data priesthood", referring to a cadre of elites who work directly with the algorithms that direct our lives. With the implication that, if we keep going the way we're going, the AIs themselves could become like a "god" overseeing everything, except that in the end it's putting money in the pockets of the surveillance capitalists.

Steo said:
Will there actually come a day where you don't get to talk to a human at all on the other side of the line anymore when you ring tech support. :shock:
Click to expand...
 
PluMGMK said:
It reminds me of Shoshana Zuboff's book on surveillance capitalism I read a few years ago. She used the term "data priesthood", referring to a cadre of elites who work directly with the algorithms that direct our lives. With the implication that, if we keep going the way we're going, the AIs themselves could become like a "god" overseeing everything, except that in the end it's putting money in the pockets of the surveillance capitalists.

Steo said:
Will there actually come a day where you don't get to talk to a human at all on the other side of the line anymore when you ring tech support. :shock:
Click to expand...
Click to expand...

I still love this show! :lol:

Hmm interesting also to imagine anything unique becoming cloned over time too, and artstyles are one of the main things which can be unique.
 
I believe Zuboff's view oversimplifies the issue of surveillance capitalism and the role of data scientists in shaping the algorithms that govern our lives. While it is true that some companies profit from the collection and analysis of personal data, and that there is a small group of individuals who have the expertise to work with these algorithms, it is important to recognize that this is just one aspect of a much larger and more complex issue.

The message implies that the rise of AI and data-driven systems is inevitable and that there is nothing that can be done to prevent it. This is not necessarily true, as there are many individuals and organizations working to address the challenges posed by surveillance capitalism and to shape the future of AI and data-driven systems in a more democratic and equitable way.

Finally, it ignores the many benefits that AI and data-driven systems can bring, such as increased efficiency, improved decision-making, and the ability to solve complex problems. While it is important to address the risks and challenges posed by surveillance capitalism, it is also important to recognize the potential benefits that these technologies can offer.
 
Well, she definitely didn't argue that nothing could be done to prevent it, just that the right moves aren't being made right now (or weren't at the time the book was written anyway).
 
I asked ChatGPT to write about RayWiki:

RayWiki is an online encyclopedia for the Rayman series of video games. It is maintained by fans of the series and provides detailed information about the characters, locations, and gameplay elements of the games. The encyclopedia is organized into articles that cover a wide range of topics related to the series, including information about the different games, characters, locations, and gameplay mechanics. In addition to the encyclopedia, RayWiki also includes a forum where users can discuss the series and share information and content related to the games.

RayWiki is a comprehensive resource for fans of the Rayman series, providing in-depth information about the games that is not available elsewhere. It is regularly updated with new content, and the community of users is active and engaged, making it a great place to connect with other fans and learn more about the series. Overall, RayWiki is a valuable resource for anyone interested in the Rayman series and is a great place to go for detailed and accurate information about the games.
Click to expand...

It doesn't quite work with Rayman Pirate-Community as a whole, but its suggestion for a happy New Year message is something I could definitely expand on:

Happy New Year to all the members of the Rayman community! We hope that the new year brings you all the joy, happiness, and gaming fun that you could wish for. Thank you for being a part of this amazing community and for sharing your love of the Rayman series with us. Here's to a great year ahead, filled with new adventures, challenges, and of course, plenty of time spent playing and enjoying the Rayman games. Wishing you all the best in the new year!
Click to expand...

:mrgreen:
 
As a theoretical/computational chemist my understanding of AI is very limited, I only know basic methods of machine learning and neural networks in order to figure out accurate parameters related to forcefields and such. AI is a completely different beast, one that I thought we were seeing the first glimpses of when ChatGPT was released before Christmas. Having spent some hours toying with it, I can only assume how many heavy filters were placed and dogwhistles were programmed that will prevent anyone from getting interesting results. Sure, I asked it some questions like "What's the nth digit of pi?" or "Give me 10 statements while only one of them is a lie", all of which were responded back to incorrectly. Not to mention that the program also just blatantly makes up stuff when it can't answer your question.

I'm absolutely convinced this program can do a ton, but because we are in the very early stages of AI right now the developers don't know how to program it in a "controlled" way yet. It's being held back and most likely for good reason considering the malicious things one could do with such a powerful tool; it's basically Tony Stark's AI Jarvis. I'm also convinced it will take uhhhh.... about 30 years maybe until we will see an AI with properties actually approaching some kind of sentience similar to humans?

Maybe more than 30 years but if I were a betting man I wouldn't say any earlier than that, even when taking the quadratic growth of such technology into account.
 
AI being used to create essays (in part or almost entirely), debate regarding what is considered "cheating", redesigning how universities teach:


I remember that back in school, the teacher would type parts of the essay in Google to see if the student just copied and pasted most of text; with AI I wonder how you can tell if a text was AI generated or human generated (specially considering that someone could fix/tweak parts of the AI generated text that seem odd, to avoid being spotted).
 
Well, I think it would be a bit suspicious if someone who doesn't write well suddenly delivers a flawless/decent essay done with AI. To be honest, it's a bit of a classist statement, but I think it's pretty accurate.

I don't know about other places, but where I work they tend to use Turnitin a lot, which is a software that allows you to detect the percentage of plagiarism in a paper (it's assumed that you're not going to do it when writing a scientific paper, but it's there just in case). I recently read that Turnitin was being improved to detect plagiarism done with AI.

www.businessinsider.com

Turnitin is the go-to software to catch students cheating. Now it's focused on a potential cat-and-mouse game with OpenAI's new ChatGPT chatbot.

The anti-plagiarism platform is rolling out a feature that can tell if a student has cheated with ChatGPT, but experts say detection will get tougher.
www.businessinsider.com www.businessinsider.com
"These models are trained at the sum of human knowledge, so they write extremely average. They are mad-lib machines that pick the most probable word in the most probable place. Humans are idiosyncratic… no person is actually exactly average"
Click to expand...

As for the way of teaching... Yes, probably quite a few things would have to be changed, but that would take a few more paragraphs to write... Maybe I should use the AI for that :fou2:
 
Finally watched both videos that have been posted here… That Pentagon stuff confused me, what exactly was going on there?

The conversation about cheating reminds me of when a History teacher in secondary school saw the word "opulent" in someone's assignment, typed the sentence into Google and found a whole load of stuff copy-pasted. He went off on a (justifiable) tirade about how we were all wasting our time if that's the sort of thing that was going on! Good to hear Turnitin is being beefed up against this…

I guess this whole story about cheating with ChatGPT also inspired this week's Dilbert comics:
dt230116.gif


Then yesterday I was talking to a colleague about a difficult customer we've had for years, since we're going to make the operating window of one of our parts more restrictive. He was like, I'm sure they'll "produce some sentences" in response to that, but we'll deal with it. I said maybe we should get ChatGPT to respond :fou:
 
PluMGMK said:
That Pentagon stuff confused me, what exactly was going on there?
Click to expand...
Yeah, I guess that was not a very good video to share. Here are some bullet points from the Daily Mail:

  • Twitter worked with the Pentagon to promote a series of fake accounts claiming to be people based in the Middle East
  • The revelation comes from the latest installment in the Twitter Files - the eighth such installment, published this time by San Francisco-based reporter Lee Fang
  • Twitter 'quietly aided the Pentagon's covert online PsyOp campaign,' Fang said, in a time frame spanning at least the years 2017-2020
  • The work was done to bolster 52 accounts tweeting in Arabic, which promoted the U.S. policies in the Middle East
  • Fang said it was done despite Twitter having 'claimed for years that they make concerted efforts to detect' and 'thwart gov-backed platform manipulation'
Click to expand...

One of the accounts that was promoted, claimed the U.S. drone bombings in Yemen killed only terrorists. The video and the article show the profile image of one of the accounts which is believed to be AI generated. I guess this is more about tech companies colluding with the government, but AI plays a part in it as a tool to deceive others.

Here is the original Twitter thread:
https://twitter.com/x/status/1605292454261182464

EDIT: Some important news I found recently (ignore the clickbait title of the video):

Artists sue Stable Diffusion maker Stability AI and others
Getty Images is suing the creators of AI art tool Stable Diffusion for scraping its content

 
Regarding plagiarism, Scott Aaronson (of quantum computing fame) is apparently working on a technique to statistically watermark ChatGPT's content by replacing the system's built-in randomness by a predetermined pseudo-random function. It's an interesting approach, and it's robust to manual tampering (e.g. replacing a couple words or inverting a couple sentences), but easily trumped by using another large language model to paraphrase the output. I feel that we can never win this game of cat and mouse and that we'll have no choice but to reconsider the way we evaluate students in the end.
 
Just a little bump. This week I went back to work as I felt a bit better of my stomach problems. I started a new project related to some iridium compounds and I had to start synthesizing these materials. Obviously, we have to look for good references with the proper software to check them out and have a clear idea about the procedure, but I was curious about what could the AI do... So I asked ChatGPT some information about what I'm doing and, surprisingly, it was very, very precise despite being a very specific area. Not onyl it gave me precise info, but also it told me precise instructions about some synthesis procedures. Here's an example.

-Tell me a synthetic route to synthesize 1-(4-(methylsulfonyl)butyl)pyrene starting up from 1-pyrene buthanol.

+A synthetic route to synthesize 1-(4-(methylsulfonyl)butyl)pyrene starting from 1-pyrene buthanol would involve several synthetic steps. One possible route could involve the following steps:

1. Protection of the hydroxyl group: The hydroxyl group on the 1-pyrene buthanol would need to be protected to prevent it from reacting in later steps. This could be accomplished by treating the compound with a protecting group such as benzyl ether or tert-butyl ether.
2. Coupling of the pyrene moiety with 4-(methylsulfonyl)butyl bromide: The pyrene moiety can be coupled with 4-(methylsulfonyl)butyl bromide using a coupling reagent such as palladium on carbon or copper(I) iodide. This reaction would likely require a base, such as sodium carbonate, to facilitate the formation of the new bond.
3. Deprotection of the hydroxyl group: Once the coupling reaction is complete, the protecting group on the hydroxyl group would need to be removed to obtain the desired 1-(4-(methylsulfonyl)butyl)pyrene. This could be accomplished by using a deprotection reagent such as hydrogen bromide (HBr) or acid.

It's important to note that the specific details of this synthesis will depend on the reagents and conditions used. Additionally, this is a general outline of one possible synthetic route and there could be other alternatives routes to obtain the same product.

It is also important to note that the synthetic route outlined above may require advanced knowledge of synthetic chemistry and experience in handling hazardous materials. It is also important to conduct this experiment under the guidance of a skilled chemist, in a well-equipped laboratory with the proper safety measures in place.
Click to expand...

This is very interesting... Will we achieve some point where the AI is so developed that we won't need all that specific software?
 
You must log in or register to reply here.
Back
Top