4 ms·

The Guardian will now use generative AI

11 points by gHeadphone 3y ago

giuliomagnifico 3y ago
> We don’t yet know the full impact that these new technologies will have on our society, but we feel sure that trusted media organisations which prioritise intelligent original reporting, uncovering facts, holding the powerful to account, and interrogating ideas will be as important as ever before I think that this explanation is good, but I also hope that the first phrase will not have a negative side, anyway experimenting with consciousness is a good thing.
jjgreen 3y ago
Chatbot gibberish in the Guardian? I guess at least the spelling will improve ...
- Ygg2 3y ago
  Oof. As occasional The Guardian reader and non-native English speaker I've seen my fair share of spelling goofs.
  - Symbiote 3y ago
    Typing errors in the Guardian are a longrunning meme. Yes Minister (1980), S01E06. "Hayward's Spinney is a vital part of Britain's heritage! Badgers have dwelt there for generations" "Well now, can you be sure of that?" "It said so in the Guardian!" "Hmm. Actually what it says here is that the 'bodgers have dealt there for generations'." https://www.dailymotion.com/video/x5v4rhe https://www.dailymotion.com/video/x5v4rhe (5:00 ish)
- zoky 3y ago
  Spelling errors? In The Grauniad?[0] Forswear! [0] https://www.theguardian.com/media/2021/may/12/guardian-200-typo-negative-best-worst-grauniad-mistakes https://www.theguardian.com/media/2021/may/12/guardian-200-t...
kumarharsh 3y ago
I guess they'll use it to write articles about how jobs are getting lost due to AI
DSingularity 3y ago
Not much will change. Models overfit on western content generating articles will not deviate from the desired western propaganda lines.
mkl 3y ago
The actual uneditorialised title is "The Guardian’s approach to generative AI", and the reality is far less sensational than the invented title "The Guardian will now use generative AI" implies.
- prox 3y ago
  Looks like we are in the next hype cycle again where people push what they want to see.
adql 3y ago
"but please continue paying us anyway" lmao, what an utter fucking disappointment of a company
helsinkiandrew 3y ago
That's a clickbait headline that isn't appropriate for the article, the original is "The Guardian’s approach to generative AI". They seem to be only using it in the background - "helping journalists interrogate large data sets, assisting colleagues through corrections or suggestions, creating ideas for marketing campaigns, or reducing the bureaucracy of time-consuming business processes" If and when they use editorially, the seem to be checking and moderating with human oversight and "will be open with readers" when they do so
- mysterydip 3y ago
  Does that mean they might do something like "hey chatgpt, what percent of people think x has the right policy on y?" and then just publish it as a factual survey?
- safety1st 3y ago
  Which is totally fine. I know several professional journalists, I even happen to employ (a former) one. None of them are using GenAI to write their articles. To help with research, topic idea and generation, maybe getting some options for how to rewrite a passage or headline, yes. All great, used right it helps them be more efficient and improves the end result. I have a very amateur background in journalism myself (edited my college and high school newspapers, took a class or two). I can smell excessive ChatGPT abuse on an article. It's like how good devs can smell bad code. ChatGPT tends to write articles which regurgitate popular viewpoints from the Internet in an equivocating way that isn't very engaging and contains no unique insights. In other words it's not output that most professional newsrooms would want to publish, it doesn't hook anybody (at least not yet). Current generative models also handle current events poorly. The Guardian are quite good at what they do and abusing e.g. ChatGPT would dilute their brand.
  - ekianjo 3y ago
    > None of them are using GenAI to write their articles. Too bad, that would probably be an improvement
vwcx 3y ago
At least they're putting forward some sort of "policy". Gannett's use is more troublesome in terms of impact: https://finance.yahoo.com/news/gannett-tiptoes-generative-ai-giving-100417356.html https://finance.yahoo.com/news/gannett-tiptoes-generative-ai...
raesene9 3y ago
Bad HN title for quite a reasonable article saying that they won't use it where there is risk of unreliable facts creeping it but will leverage it for things like interrogating large datasets. Makes sense to me, generating queries using natural language seems to be one of the less tricky areas to get right.
Topfi 3y ago
This honestly seems like a reasonable, sober approach to LLMs that considers how they may serve journalists in a few key ways, whilst retaining them as the ones actually writing the content. Using LLMs to help sift through large datasets and assist the (hopefully sufficiently Church/State separated) marketing team is honestly a great way of staying able to deliver accurate information to readers. The title (for some reason editorialized from the more fitting "The Guardian’s approach to generative AI") could give a person that doesn't read their release the wrong idea that they may consider utilizing LLMs for article creation in a way that they at this point in time don't seem to intend. Don't know why the change from the original was made, as this can only lead to confusion and controversy and does not serve anyone (unlike e.g. adding the year something was posted).

11 points by gHeadphone 3y ago

> We don’t yet know the full impact that these new technologies will have on our society, but we feel sure that trusted media organisations which prioritise intelligent original reporting, uncovering facts, holding the powerful to account, and interrogating ideas will be as important as ever before I think that this explanation is good, but I also hope that the first phrase will not have a negative side, anyway experimenting with consciousness is a good thing.

jjgreen 3y ago

Chatbot gibberish in the Guardian? I guess at least the spelling will improve ...

Ygg2 3y ago
Oof. As occasional The Guardian reader and non-native English speaker I've seen my fair share of spelling goofs.
- Symbiote 3y ago
  Typing errors in the Guardian are a longrunning meme. Yes Minister (1980), S01E06. "Hayward's Spinney is a vital part of Britain's heritage! Badgers have dwelt there for generations" "Well now, can you be sure of that?" "It said so in the Guardian!" "Hmm. Actually what it says here is that the 'bodgers have dealt there for generations'." https://www.dailymotion.com/video/x5v4rhe https://www.dailymotion.com/video/x5v4rhe (5:00 ish)
zoky 3y ago
Spelling errors? In The Grauniad?[0] Forswear! [0] https://www.theguardian.com/media/2021/may/12/guardian-200-typo-negative-best-worst-grauniad-mistakes https://www.theguardian.com/media/2021/may/12/guardian-200-t...

kumarharsh 3y ago

I guess they'll use it to write articles about how jobs are getting lost due to AI

DSingularity 3y ago

Not much will change. Models overfit on western content generating articles will not deviate from the desired western propaganda lines.

mkl 3y ago

The actual uneditorialised title is "The Guardian’s approach to generative AI", and the reality is far less sensational than the invented title "The Guardian will now use generative AI" implies.

prox 3y ago
Looks like we are in the next hype cycle again where people push what they want to see.

adql 3y ago

"but please continue paying us anyway" lmao, what an utter fucking disappointment of a company

helsinkiandrew 3y ago

That's a clickbait headline that isn't appropriate for the article, the original is "The Guardian’s approach to generative AI". They seem to be only using it in the background - "helping journalists interrogate large data sets, assisting colleagues through corrections or suggestions, creating ideas for marketing campaigns, or reducing the bureaucracy of time-consuming business processes" If and when they use editorially, the seem to be checking and moderating with human oversight and "will be open with readers" when they do so

mysterydip 3y ago
Does that mean they might do something like "hey chatgpt, what percent of people think x has the right policy on y?" and then just publish it as a factual survey?
safety1st 3y ago
Which is totally fine. I know several professional journalists, I even happen to employ (a former) one. None of them are using GenAI to write their articles. To help with research, topic idea and generation, maybe getting some options for how to rewrite a passage or headline, yes. All great, used right it helps them be more efficient and improves the end result. I have a very amateur background in journalism myself (edited my college and high school newspapers, took a class or two). I can smell excessive ChatGPT abuse on an article. It's like how good devs can smell bad code. ChatGPT tends to write articles which regurgitate popular viewpoints from the Internet in an equivocating way that isn't very engaging and contains no unique insights. In other words it's not output that most professional newsrooms would want to publish, it doesn't hook anybody (at least not yet). Current generative models also handle current events poorly. The Guardian are quite good at what they do and abusing e.g. ChatGPT would dilute their brand.
- ekianjo 3y ago
  > None of them are using GenAI to write their articles. Too bad, that would probably be an improvement

vwcx 3y ago

At least they're putting forward some sort of "policy". Gannett's use is more troublesome in terms of impact: https://finance.yahoo.com/news/gannett-tiptoes-generative-ai-giving-100417356.html https://finance.yahoo.com/news/gannett-tiptoes-generative-ai...

raesene9 3y ago

Bad HN title for quite a reasonable article saying that they won't use it where there is risk of unreliable facts creeping it but will leverage it for things like interrogating large datasets. Makes sense to me, generating queries using natural language seems to be one of the less tricky areas to get right.

Topfi 3y ago

This honestly seems like a reasonable, sober approach to LLMs that considers how they may serve journalists in a few key ways, whilst retaining them as the ones actually writing the content. Using LLMs to help sift through large datasets and assist the (hopefully sufficiently Church/State separated) marketing team is honestly a great way of staying able to deliver accurate information to readers. The title (for some reason editorialized from the more fitting "The Guardian’s approach to generative AI") could give a person that doesn't read their release the wrong idea that they may consider utilizing LLMs for article creation in a way that they at this point in time don't seem to intend. Don't know why the change from the original was made, as this can only lead to confusion and controversy and does not serve anyone (unlike e.g. adding the year something was posted).