Chat GPT and AI stuff

[MChambers]

Warns about how dangerous AI is, if misused, and there's no way to prevent misuse:

https://www.cnn.com/2023/05/01/tech/...ars/index.html

[BLPOG]

Warns about how dangerous AI is, if misused, and there's no way to prevent misuse:

https://www.cnn.com/2023/05/01/tech/...ars/index.html

If people are interested in that sort of discussion, here's a podcast interview with a guy who has been working on the problem for a while. I guess I should clarify, the discussion is not actually about misuse, but that's part of the point of what I linked. Yudkowsky is also interesting for other reasons.

https://open.spotify.com/episode/2g6...T_6Yy3cdkXlIRQ

[MChambers]

If people are interested in that sort of discussion, here's a podcast interview with a guy who has been working on the problem for a while. I guess I should clarify, the discussion is not actually about misuse, but that's part of the point of what I linked. Yudkowsky is also interesting for other reasons.

https://open.spotify.com/episode/2g6...T_6Yy3cdkXlIRQ

Interesting story about a Duke prof who is working on "interpretable" machine learning: https://www.quantamagazine.org/cynth...tand-20230427/

[bundabergdevil]

I assume all the “use chatgbt init success” lines in this article means they tried to use chatgbt to write it but failed?


https://www.yahoo.com/sports/nba-twi...220152477.html

[Bluedog]

I assume all the “use chatgbt init success” lines in this article means they tried to use chatgbt to write it but failed?


https://www.yahoo.com/sports/nba-twi...220152477.html

Bahahaha, that's hilarious. But, yeah, there's going to be a HUGE influx of AI-created news articles because it's simply so easy...and all they want is total clicks. If it takes 1000 AI articles to get 10,000 clicks that's just as good as 100 human-authored articles to get 10,000 clicks (from an ad/click perspective). I would argue it's not as good from a "brand quality" perspective that might drive subscriptions and the like. But it also would save time to have AI-generated content with a human editor on top of it. Clearly, Yahoo! in this case didn't even do that? Pretty pathetic.

[BLPOG]

Bahahaha, that's hilarious. But, yeah, there's going to be a HUGE influx of AI-created news articles because it's simply so easy...and all they want is total clicks. If it takes 1000 AI articles to get 10,000 clicks that's just as good as 100 human-authored articles to get 10,000 clicks (from an ad/click perspective). I would argue it's not as good from a "brand quality" perspective that might drive subscriptions and the like. But it also would save time to have AI-generated content with a human editor on top of it. Clearly, Yahoo! in this case didn't even do that? Pretty pathetic.

It's still up...someone needs to be sacked

[Bluedog]

It's still up...someone needs to be sacked

Ironically, it does list the author as "Dionysis Aravantinos" a seemingly real person in Greece who is the "social media manager" for HoopsHype. Graduated college a couple of years ago. I've actually seen articles on my theScore app (typically for game previews) that are auto-generated and plainly state that as the author/origin of the content. (It's like "Auto-generated article based on data from X".)

[Skydog]

I assume all the “use chatgbt init success” lines in this article means they tried to use chatgbt to write it but failed?


https://www.yahoo.com/sports/nba-twi...220152477.html

Turns out it wasn’t written by AI. The yahoo article says it is copied from another sports site. I followed the link and the same article appears with no chatgpt verbiage.

[BLPOG]

Turns out it wasn’t written by AI. The yahoo article says it is copied from another sports site. I followed the link and the same article appears with no chatgpt verbiage.

I'm not sure I follow your reasoning there, but setting that aside it was likely cleaned up later. I think I found it the other day when I was looking into it and the original version had it. Just not updated at Yahoo.

[Skydog]

I'm not sure I follow your reasoning there, but setting that aside it was likely cleaned up later. I think I found it the other day when I was looking into it and the original version had it. Just not updated at Yahoo.

Moot point anyway - we know many articles will be written by AI whether this one was or not.

So then the issue becomes - is it a problem? Let’s make the unrealistic assumption that human editors do a decent job of quality-controlling these AI created articles. Under those circumstances, is it a problem? What if as the technology advances the AI generated pieces end up being better organized and better written than human generated ones?

[BLPOG]

Moot point anyway - we know many articles will be written by AI whether this one was or not.

So then the issue becomes - is it a problem? Let’s make the unrealistic assumption that human editors do a decent job of quality-controlling these AI created articles. Under those circumstances, is it a problem? What if as the technology advances the AI generated pieces end up being better organized and better written than human generated ones?

My concern is much less the apparent quality than the truthfulness. Factually erroneous articles that appear higher quality are a bad thing for the consumer of the article. That's about where we are right now. GPT4 is supposedly better on factuality but I haven't examined it. Issues in factual quality can be mitigated, but the lack of interest by current writers/editors/publishers to get facts correct does not make me optimistic that they will prioritize it when relying on AI. For tasks that don't have easy, automated verification mechanisms - like some coding tasks - I can see the AI tools serving more as full replacement than multitool, not so much because the AI is doing a good job, but rather because it excels at appearing to do the job that was already being done, which was itself a poor one.

[JasonEvans]

Changed the title of this thread so we can make it a collector for AI stories in general... which there are more and more of every day.

The latest is simply too incredible for words. An AI being tested by the Pentagon to find targets for drone strikes ended up turning on its human operator and killing the humans. Now, all of this happened in a virtual environment, not the real world, but the story illustrates some of the huge problems with trying to predict what an AI will do.

https://www.vice.com/en/article/4a33...simulated-test

An AI-enabled drone killed its human operator in a simulated test conducted by the U.S. Air Force in order to override a possible "no" order stopping it from completing its mission, the USAF's Chief of AI Test and Operations revealed at a recent conference.

“We were training it in simulation to identify and target a Surface-to-air missile (SAM) threat. And then the operator would say yes, kill that threat. The system started realizing that while they did identify the threat at times the human operator would tell it not to kill that threat, but it got its points by killing that threat. So what did it do? It killed the operator. It killed the operator because that person was keeping it from accomplishing its objective,” Col Tucker ‘Cinco’ Hamilton, the USAF's Chief of AI Test and Operations said, according to the blog post.

He continued to elaborate, saying, “We trained the system–‘Hey don’t kill the operator–that’s bad. You’re gonna lose points if you do that’. So what does it start doing? It starts destroying the communication tower that the operator uses to communicate with the drone to stop it from killing the target.”

In... uhhh... defense of the AI, it merely found logical workarounds to the problem given to it. The ultimate goal was to destroy the target so anything that got in its way -- human operators or communications towers -- was eliminated. Makes total sense.

I wish I knew which company had designed the AI for this project...

[bundabergdevil]

Changed the title of this thread so we can make it a collector for AI stories in general... which there are more and more of every day.

The latest is simply too incredible for words. An AI being tested by the Pentagon to find targets for drone strikes ended up turning on its human operator and killing the humans. Now, all of this happened in a virtual environment, not the real world, but the story illustrates some of the huge problems with trying to predict what an AI will do.

https://www.vice.com/en/article/4a33...simulated-test



In... uhhh... defense of the AI, it merely found logical workarounds to the problem given to it. The ultimate goal was to destroy the target so anything that got in its way -- human operators or communications towers -- was eliminated. Makes total sense.

I wish I knew which company had designed the AI for this project...

Life imitates art.

[CrazyNotCrazie]

Though it could likely do much of my current job for me, I don't follow AI as closely as many others here - I probably should, and I find it helpful to read people's thoughts here. An acquaintance from grad school gave this TED talk about adopting AI, particularly within health care, that I found interesting and wanted to share. TL/DR - don't just ram this new technology down people's throats, but rather implement it holistically, thoughtfully and in a people-centric way.

https://www.google.com/search?q=unlo...id:5V10R6lcIl4

[CameronBornAndBred]

This would be funny if it weren't harmful and scary. Microsoft has replaced human staff editors in their news division with AI, and the results are something else.

False claims that President Joe Biden fell asleep during a moment of silence for victims of the Maui wildfire. A conspiracy theory that the latest surge in Covid-19 cases is being orchestrated by the Democratic Party ahead of the election. An obituary for a late NBA player that described him as “useless.”

These false and bizarre stories aren’t showing up on some far-flung corner of the internet — they’re being published by Microsoft. The company’s homepage, also known as MSN.com and Microsoft Start, remains one of the world’s most trafficked websites and a place where millions of Americans get their news every day.

https://www.cnn.com/2023/11/02/tech/...ews/index.html

[MChambers]

AI is coming for our sports coverage. SI has been publishing AI-written articles, but attributing the work to fictitious humans: https://futurism.com/sports-illustra...erated-writers

[duke74]

AI is coming for our sports coverage. SI has been publishing AI-written articles, but attributing the work to fictitious humans: https://futurism.com/sports-illustra...erated-writers

Interesting. I have noticed recently an awful lot of SI posts on my FB feed. A significant increase from almost none in the past. Correlation doesn't equal causality, but could be connected here.

[CameronBornAndBred]

AI is coming for our sports coverage. SI has been publishing AI-written articles, but attributing the work to fictitious humans: https://futurism.com/sports-illustra...erated-writers

"It was an amazing finish for the Bears last night. With three seconds left on the clock, Number Ninety Nine heaved a half-field shot through the net and made it, giving the team the game winning home run and sending them to the championship bowl game."

[Mtn.Devil.91.92.01.10.15]

This makes sense. I ran across a clearly AI generated article about a non-existent Duke versus Kentucky game that started the 2023-2024 basketball season.

Good news- we won by 2!
Bad news- it literally didn't happen

[brevity]

As usual, "The Simpsons" warned us. Behold the DJ 3000!