Post ASo6p6k0KZ5c1s8vTM by rmkelly@zirk.us
(DIR) More posts by rmkelly@zirk.us
(DIR) Post #ASm5dsCnXqHjUZloxs by grammargirl@zirk.us
2023-02-17T15:30:49Z
0 likes, 0 repeats
This story about Bing's #chatGPT has some of the most jaw-dropping examples of inappropriate responses I've seen. A few are so outrageous they're funny, but this is not: “It can suggest that users harm others,” he said. “These are far more serious issues than the tone being off.” 1/2https://apnews.com/article/technology-science-microsoft-corp-business-software-fb49e5d625bf37be0527e5173116bef3
(DIR) Post #ASm5pJIJAYy4gynjUG by grammargirl@zirk.us
2023-02-17T15:32:57Z
0 likes, 0 repeats
I also see why researchers are losing their minds about the way the media is portraying these systems as sentient."Microsoft declined further comment about Bing’s behavior Thursday, but Bing itself agreed to comment..."It didn't "agree to comment." They fed it a prompt, and it responded as it is programmed to do.
(DIR) Post #ASm5yVlQB2wJyKo208 by StaceyCornelius@mstdn.ca
2023-02-17T15:34:37Z
0 likes, 0 repeats
@grammargirl What? Scraping the web results in terrible responses? Inconceivable!(sigh)
(DIR) Post #ASm6JFDWtNwqjcq7iS by jrod3737@mstdn.social
2023-02-17T15:38:22Z
0 likes, 0 repeats
@grammargirl Wait, a non-sentient computer program with unchecked autonomy did something we didn’t want it to do?Pretends to be shocked
(DIR) Post #ASm6eOmGXhhuT0AsN6 by Jantar@mstdn.social
2023-02-17T15:42:07Z
0 likes, 0 repeats
@grammargirl That is going well. Still, they're only at the B stage. By the time they reach the S, things will have improved considerably, I'm sure - so let's not be too negative, until we've heard version Santos doing its thing.
(DIR) Post #ASm6nByVVmKRMn5qcK by BudGibson@mstdn.social
2023-02-17T15:43:47Z
0 likes, 0 repeats
@grammargirl In a very colloquial sense, it did agree to comment since it generated an answer to a question that corporate decided not to respond to. Microsoft has placed Bing ChatGPT before us as a responder. It has no way of telling it not to respond.While I do get your point, I see no reason for people not to just use their usual social heuristics in judging conversations. Probably, the wisest route is to just not engage when an actor is violating those norms.
(DIR) Post #ASm77M5pbOuoXU8LZo by iDGS@mas.to
2023-02-17T15:47:22Z
0 likes, 0 repeats
@grammargirl See 🧵by @sundress. First post: “I keep reading things written about generative AI that point to folks not having read the titular paragraph of the Stochastic Parrots paper: "Text generated by an LM is not grounded in communicative intent, any model of the world, or any model of the reader’s state of mind. It can’t have been, because the training data never included sharing thoughts with a listener, nor does the machine have the ability to do that.(continued)”A “must read!”
(DIR) Post #ASm7iaicBhJOc81uSW by grammargirl@zirk.us
2023-02-17T15:54:03Z
0 likes, 0 repeats
@BudGibson I believer researchers (at least some) believe it actually is dangerous to anthropomorphize AI models because doing so changes the way the average user perceives it and interacts with it.
(DIR) Post #ASm8Rk8c0uDwrIih4y by BudGibson@mstdn.social
2023-02-17T16:02:15Z
0 likes, 0 repeats
@grammargirl Thanks, I think this correctly puts the onus on the model builders. There is a whole field of AI ethics that waxes and wanes. In the courses I teach, students use such assistants all the time to create ads, but the ad components are presented as options.
(DIR) Post #ASmAGfCwCgYWI5aVIu by stoklund@oldbytes.space
2023-02-17T16:22:05Z
0 likes, 0 repeats
@grammargirl this is pretty much the same thing that happened with ELIZA in the sixties https://en.wikipedia.org/wiki/ELIZA
(DIR) Post #ASmAINMVWZC5WSsdw8 by kyleha@mastodon.social
2023-02-17T16:23:01Z
0 likes, 0 repeats
@grammargirl Yes! It's a thing, not a human!I could write a program to randomly select words from a dictionary to produce nonsense. It has exactly the same kind of agency and intent as Bing. The "AI" just has a bigger dictionary and a more complicated selection algorithm.
(DIR) Post #ASmNzxEdK8redYJ7eC by grammargirl@zirk.us
2023-02-17T18:56:27Z
0 likes, 0 repeats
@stoklund Fascinating. Thanks!
(DIR) Post #ASmSeKtx9SIzRA94mO by NicoleCRust@neuromatch.social
2023-02-17T19:48:41Z
0 likes, 0 repeats
@grammargirl Words matter. ⬆️
(DIR) Post #ASmUo5TRZ7eiOSWVJg by ordinal@mas.to
2023-02-17T20:12:49Z
0 likes, 0 repeats
@grammargirl this piece by the Centre for Privacy & Technology is one I keep coming back to https://medium.com/center-on-privacy-technology/artifice-and-intelligence%C2%B9-f00da128d3cd
(DIR) Post #ASmrkRrgfjctcq64OG by grammargirl@zirk.us
2023-02-18T00:29:56Z
0 likes, 0 repeats
@ordinal That was really interesting. Thank you!
(DIR) Post #ASnBKM6rRh7xKW1u1g by bizcad@dotnet.social
2023-02-18T04:09:18Z
0 likes, 0 repeats
@grammargirl @NicoleCRust I recently read this very good article on what "AI" is really doing.https://tinyurl.com/yedzj4ybStephen Wolfram does a very creditable job of demystifying how LLMs work.
(DIR) Post #ASo6F6Fgfgci5UnCL2 by wilson@nfld.me
2023-02-18T14:46:51Z
0 likes, 0 repeats
@grammargirl
(DIR) Post #ASo6p6k0KZ5c1s8vTM by rmkelly@zirk.us
2023-02-18T14:53:28Z
0 likes, 0 repeats
@grammargirl Exactly. The journalists are being fed a line by the likes of Microsoft's Kevin Scott, and many of the journalists, who are supposed to know better, are swallowing it whole.