7 ms·

I think what you're saying is valid, but is only one angle of looking at it. LLMs, in my opinion, are on average better at writing copy than humans. Or rather

I think what you're saying is valid, but is only one angle of looking at it. LLMs, in my opinion, are on average better at writing copy than humans. Or rather they manage to do so more consistently. Thus, even if humans still need to provide supervision, the clarity and conciseness of government documents may be improved overall.

pclmulqdq 3y ago
Having seen others use GPT-4 for a bunch of business documents, I disagree. Its advantage is that it produces average-- (40-50th percentile) copy very quickly, but that quality degrades as you ask for documents that are more unique (ie not in the training set). Incidentally, this makes a fine tuned version perfect for the government, which produces a shit load of low-effort documents, but today employs a bunch of morons to write them. If you want a lot of text that you don't want to think about, LLMs are great. If you want text that is pithy or persuasive, it doesn't help much, even when you use prompting tricks (eg "you are a CEO/professor..."). By the way, this has convinced me that local LLAMA-scale models are the future, not massive remote GPTs.
- hdlothia 3y ago
  Implying all government employees are morons is pretty harsh. I know a lot of intelligent, driven people working in government or for government contractors. I know we think we're gods here for writing websites but a little humility goes a long way.
  - pclmulqdq 3y ago
    I know a lot of government employees who are smart and driven, too. They are not the people who write things like IRS audit letters, military after-action reports, and the like. Many government employees are barely literate, and told to spend all day writing stuff.
    - hdlothia 3y ago
      I'll have to agree to disagree. The government certainly employs a solid amount of stupid people, as does every institution, but I would quibble with the idea that writing reports as your main job function makes you more likely to be a barely literate moron. I appreciate the rest of your post though and I think you make a great point about AI being very effective for certain types of work
      - pclmulqdq 3y ago
        For the record, I didn't say that writing reports all day makes you a barely literate moron. I said that a lot of barely literate morons are told to write reports all day - very different - and categorized the kind of reports that these folks often write. Lots of smart people also write reports all day at the government, and probably would also benefit from a tool that makes their writing a lot faster but worse, since the quality basically doesn't matter.
    - whompyjaw 3y ago
      Do you know this for certain? As in, have you done the analysis to prove this? Collected all the IQ and other aptitude tests of these people to conclude this for certain? Some of them could be in the beginning part of their government career attending school. Did you run a survey to ask these people what they do with their lives?
- ravenstine 3y ago
  Yeah, I'm sure it depends on the training and the document type. I haven't had direct experience with LLM-generated business documents, but I do think that, given a small time window, GPT-4 is more effective at communicating ideas than the average human even if those ideas are wrong. As far as my use goes, GPT-4 now serves the role that Wikipedia used to, which is to give me a baseline of understanding that I use to refine my research, given that most internet content is either verbose, hogwash, or both. Ironically, I've found that Wikipedia has made itself nearly obsolete in this regard given how seemingly every article has become so academically-written that it's an undertaking to just get a rough idea of something. Neither source I trust, but at least GPT-4 will introduce me to a topic in a way that is more effective and concise than most human writing today. But that's just my experience. > If you want a lot of text that you don't want to think about, LLMs are great. If you want text that is pithy or persuasive, it doesn't help much, even when you use prompting tricks (eg "you are a CEO/professor..."). By the way, this has convinced me that local LLAMA-scale models are the future, not massive remote GPTs. Granted, we are judging LLMs based on highly generalized training sets. What if a GPT was fine-tuned on all of the writing and speeches of figures whom are considered the most persuasive?
- mymythisisthis 3y ago
  A bleak Byzantine future, of endless bureaucracy. Government reports continue to balloon in size, because of technological advancements (especially copy/paste). In the age of typewriters, anything not important, or duplicate, was cut. With word processors, document sizes unnecessarily grew.
arp242 3y ago
On the other hand reviewing and correcting text is often harder and requires more attention than writing it yourself: it's easy to miss something, and just assume "ah I think that's fine" (when it's actually not). All things considered, an awkward sentence or two isn't really that much of a big deal compared to an incorrect assertion that was missed in review. I also don't think GPT-4 is necessarily that brilliant at writing copy, but I don't really know how good/bad US government officials are so I can't really compare, but in other parts of the world I haven't really that much of an issue with it.
lm28469 3y ago
I'll never get this argument. If all you do is give bullet lists to chatgpt to get 50 pages of elaborated text just give me the damn bullet list. We'll both save time and CPU cycles For example right now we have AIs writing CVs and cover letters that are read by AIs, we're "optimizing" things that shouldn't exist anymore but we're too deep in the cycle to even notice it
- pvarangot 3y ago
  He probably doesn't want you to see the bullet list because it gives up his agenda.
- mistermann 3y ago
  > I'll never get this argument. If all you do is give bullet lists to chatgpt to get 50 pages of elaborated text just give me the damn bullet list. We'll both save time and CPU cycles Some communication is intended to inform, some is intended to persuade... And often, the speaker doesn't even really know which they are dealing with.
- renewiltord 3y ago
  Likely your stated and revealed preferences are different. A place I noticed this was tech people (who are not in ad-tech) talking about advertisements: "Just describe the thing to me and if I like the specs and functionality, I'll buy it" but then when they go to buy the thing it's more of the thing that's not doing that and any time anything does that it's "oh that's cool but if it did X" but then the thing they end up buying won't do X either. The best way to get people to interact with you a certain way is to reward them for interacting with you that way, and you likely don't. Not because that's a failure of you but it's just how people (all of us) act. The world isn't a co-operative multi-tasking OS. We're more like the human body: a multi-agent conglomerate whose component organisms communicate via signals that are filtered and processed. Sometimes a pathogen mimics signals well enough. Other times we detect it. But the system wouldn't work if every participant trusted the counterparty.
- JohnFen 3y ago
  Yes, this is what I've been wondering, too. For uses like this, it seems that just publishing the prompt itself would be more efficient than giving the output the prompt generates.
fzeindl 3y ago
> LLMs, in my opinion, are on average better at writing copy than humans. I disagree here. But what I found LLMs really useful for is summarizing topics like: "Suppose I want to do XYZ, which steps should I take?" or "I want to write about ABC, how would you structure the article?" The content is not very good if you look at it in detail, but LLMs provide good overviews and gave me ideas on parts of a topic I overlooked. More librarian than search engine. I cannot copy and paste and rely on the results but I can rely on that they point me in a good direction.
- ravenstine 3y ago
  What you describe is one of the jobs that most text is currently failing at. Writing isn't just about documenting every itty bitty detail in academic verbosity, but to introduce people to ideas using effective summaries. The vast majority of what's out there flat out fails at this for at least a few reasons; many people have a very scatterbrained writing style to begin with, many authoritative sources provide more verbosity than necessary for the target audience, and the rest is often SEO garbage that speaks for itself. > More librarian than search engine. I cannot copy and paste and rely on the results but I can rely on that they point me in a good direction. The same can be said of physical books, and most content should be written more like that in books than what the internet is currently selecting for.
damnesian 3y ago
I don't quite agree with this. What GPT does is extraordinary for a machine, but it is not terribly skilled compared to a human trained to be a fully capable communicator. What it primarily does is save the time humans would need to fulfill the brief. It automates the basics. That does not automatically elevate it to excellent or even beyond merely competent. For the boilerplate- well sure, but I could do that with existing tools, no AI needed. Custom communications I'm sure AI can handle the very basics. but once a human replies and consideration is needed for special cases- can you imagine ChatGPT counseling you on a healthcare decision? I can't. I can't imagine ChatGPT doing much more than being an autoattendant until a human can reply fully, except the message humans can send to the agency will be customized and drafted according to a standard intake process.
indymike 3y ago
> LLMs, in my opinion, are on average better at writing copy than humans. The question is, do you want an AI writing the brief and doing the research that is going to be given to an administrator who is going to decide if you are denied medical care in an emergency?
bamfly 3y ago
> LLMs, in my opinion, are on average better at writing copy than humans. A disturbingly high proportion of US adults (at least in the US—I don't know how this looks in other countries) barely count as literate. Perhaps half[1]. Most of the rest aren't a ton better, and can't write worth a damn. I think "on average" understates the case: at least 90% of US adults would probably get better results by passing their writing through ChatGPT, than not. Most of that remaining 10% would also see improvements from ChatGPT, except for writing they've put a great deal of effort into. It's very good at cleaning up writing, and/or tuning it to be better-suited to some purpose than the original was. In fact, a large majority of readers in the US would probably benefit from passing what they read through ChatGPT and having it summarize it concisely, in very simple terms. The world makes a lot more sense, to those who grew up taking to the written word as naturally as a duck to water, when we appreciate that the world runs on the written word, yet most people are terrible readers and worse writers. [1] "According to a 2020 report by the U.S. Department of Education, 54% of adults in the United States have prose literacy below the 6th-grade level." And, nb, if you're a good reader and you're thinking "6th grade level, that's not so bad", your idea of what a "6th-grade level" looks like, is probably skewed way higher than what they mean. https://en.wikipedia.org/wiki/Literacy_in_the_United_States https://en.wikipedia.org/wiki/Literacy_in_the_United_States
- yyyk 3y ago
  >at least 90% of US adults would probably get better results by passing their writing through ChatGPT, than not.... In fact, a large majority of readers in the US would probably benefit from passing what they read through ChatGPT Bullet list -> AI text generator -> AI text summarizer -> Bullet list Perhaps OpenAI can optimize the process by storing the original prompt and just giving it back at the last phase.
- stOneskull 3y ago
  > 54% of adults in the United States have prose literacy below the 6th-grade level. this is quite saddening.
  - bamfly 3y ago
burnished 3y ago
I havent seen a compelling example of that, to the point that I think it has associated 'intelligence' with verbosity. I'd expect writing samples to get shorter, not longer, when being improved.

pclmulqdq 3y ago

Having seen others use GPT-4 for a bunch of business documents, I disagree. Its advantage is that it produces average-- (40-50th percentile) copy very quickly, but that quality degrades as you ask for documents that are more unique (ie not in the training set). Incidentally, this makes a fine tuned version perfect for the government, which produces a shit load of low-effort documents, but today employs a bunch of morons to write them. If you want a lot of text that you don't want to think about, LLMs are great. If you want text that is pithy or persuasive, it doesn't help much, even when you use prompting tricks (eg "you are a CEO/professor..."). By the way, this has convinced me that local LLAMA-scale models are the future, not massive remote GPTs.

hdlothia 3y ago
Implying all government employees are morons is pretty harsh. I know a lot of intelligent, driven people working in government or for government contractors. I know we think we're gods here for writing websites but a little humility goes a long way.
- pclmulqdq 3y ago
  I know a lot of government employees who are smart and driven, too. They are not the people who write things like IRS audit letters, military after-action reports, and the like. Many government employees are barely literate, and told to spend all day writing stuff.
  - hdlothia 3y ago
    I'll have to agree to disagree. The government certainly employs a solid amount of stupid people, as does every institution, but I would quibble with the idea that writing reports as your main job function makes you more likely to be a barely literate moron. I appreciate the rest of your post though and I think you make a great point about AI being very effective for certain types of work
    - pclmulqdq 3y ago
      For the record, I didn't say that writing reports all day makes you a barely literate moron. I said that a lot of barely literate morons are told to write reports all day - very different - and categorized the kind of reports that these folks often write. Lots of smart people also write reports all day at the government, and probably would also benefit from a tool that makes their writing a lot faster but worse, since the quality basically doesn't matter.
  - whompyjaw 3y ago
    Do you know this for certain? As in, have you done the analysis to prove this? Collected all the IQ and other aptitude tests of these people to conclude this for certain? Some of them could be in the beginning part of their government career attending school. Did you run a survey to ask these people what they do with their lives?
ravenstine 3y ago
Yeah, I'm sure it depends on the training and the document type. I haven't had direct experience with LLM-generated business documents, but I do think that, given a small time window, GPT-4 is more effective at communicating ideas than the average human even if those ideas are wrong. As far as my use goes, GPT-4 now serves the role that Wikipedia used to, which is to give me a baseline of understanding that I use to refine my research, given that most internet content is either verbose, hogwash, or both. Ironically, I've found that Wikipedia has made itself nearly obsolete in this regard given how seemingly every article has become so academically-written that it's an undertaking to just get a rough idea of something. Neither source I trust, but at least GPT-4 will introduce me to a topic in a way that is more effective and concise than most human writing today. But that's just my experience. > If you want a lot of text that you don't want to think about, LLMs are great. If you want text that is pithy or persuasive, it doesn't help much, even when you use prompting tricks (eg "you are a CEO/professor..."). By the way, this has convinced me that local LLAMA-scale models are the future, not massive remote GPTs. Granted, we are judging LLMs based on highly generalized training sets. What if a GPT was fine-tuned on all of the writing and speeches of figures whom are considered the most persuasive?
mymythisisthis 3y ago
A bleak Byzantine future, of endless bureaucracy. Government reports continue to balloon in size, because of technological advancements (especially copy/paste). In the age of typewriters, anything not important, or duplicate, was cut. With word processors, document sizes unnecessarily grew.

arp242 3y ago

On the other hand reviewing and correcting text is often harder and requires more attention than writing it yourself: it's easy to miss something, and just assume "ah I think that's fine" (when it's actually not). All things considered, an awkward sentence or two isn't really that much of a big deal compared to an incorrect assertion that was missed in review. I also don't think GPT-4 is necessarily that brilliant at writing copy, but I don't really know how good/bad US government officials are so I can't really compare, but in other parts of the world I haven't really that much of an issue with it.

lm28469 3y ago

I'll never get this argument. If all you do is give bullet lists to chatgpt to get 50 pages of elaborated text just give me the damn bullet list. We'll both save time and CPU cycles For example right now we have AIs writing CVs and cover letters that are read by AIs, we're "optimizing" things that shouldn't exist anymore but we're too deep in the cycle to even notice it

pvarangot 3y ago
He probably doesn't want you to see the bullet list because it gives up his agenda.
mistermann 3y ago
> I'll never get this argument. If all you do is give bullet lists to chatgpt to get 50 pages of elaborated text just give me the damn bullet list. We'll both save time and CPU cycles Some communication is intended to inform, some is intended to persuade... And often, the speaker doesn't even really know which they are dealing with.
renewiltord 3y ago
Likely your stated and revealed preferences are different. A place I noticed this was tech people (who are not in ad-tech) talking about advertisements: "Just describe the thing to me and if I like the specs and functionality, I'll buy it" but then when they go to buy the thing it's more of the thing that's not doing that and any time anything does that it's "oh that's cool but if it did X" but then the thing they end up buying won't do X either. The best way to get people to interact with you a certain way is to reward them for interacting with you that way, and you likely don't. Not because that's a failure of you but it's just how people (all of us) act. The world isn't a co-operative multi-tasking OS. We're more like the human body: a multi-agent conglomerate whose component organisms communicate via signals that are filtered and processed. Sometimes a pathogen mimics signals well enough. Other times we detect it. But the system wouldn't work if every participant trusted the counterparty.
JohnFen 3y ago
Yes, this is what I've been wondering, too. For uses like this, it seems that just publishing the prompt itself would be more efficient than giving the output the prompt generates.

fzeindl 3y ago

> LLMs, in my opinion, are on average better at writing copy than humans. I disagree here. But what I found LLMs really useful for is summarizing topics like: "Suppose I want to do XYZ, which steps should I take?" or "I want to write about ABC, how would you structure the article?" The content is not very good if you look at it in detail, but LLMs provide good overviews and gave me ideas on parts of a topic I overlooked. More librarian than search engine. I cannot copy and paste and rely on the results but I can rely on that they point me in a good direction.

ravenstine 3y ago
What you describe is one of the jobs that most text is currently failing at. Writing isn't just about documenting every itty bitty detail in academic verbosity, but to introduce people to ideas using effective summaries. The vast majority of what's out there flat out fails at this for at least a few reasons; many people have a very scatterbrained writing style to begin with, many authoritative sources provide more verbosity than necessary for the target audience, and the rest is often SEO garbage that speaks for itself. > More librarian than search engine. I cannot copy and paste and rely on the results but I can rely on that they point me in a good direction. The same can be said of physical books, and most content should be written more like that in books than what the internet is currently selecting for.

damnesian 3y ago

I don't quite agree with this. What GPT does is extraordinary for a machine, but it is not terribly skilled compared to a human trained to be a fully capable communicator. What it primarily does is save the time humans would need to fulfill the brief. It automates the basics. That does not automatically elevate it to excellent or even beyond merely competent. For the boilerplate- well sure, but I could do that with existing tools, no AI needed. Custom communications I'm sure AI can handle the very basics. but once a human replies and consideration is needed for special cases- can you imagine ChatGPT counseling you on a healthcare decision? I can't. I can't imagine ChatGPT doing much more than being an autoattendant until a human can reply fully, except the message humans can send to the agency will be customized and drafted according to a standard intake process.

indymike 3y ago

> LLMs, in my opinion, are on average better at writing copy than humans. The question is, do you want an AI writing the brief and doing the research that is going to be given to an administrator who is going to decide if you are denied medical care in an emergency?

bamfly 3y ago

> LLMs, in my opinion, are on average better at writing copy than humans. A disturbingly high proportion of US adults (at least in the US—I don't know how this looks in other countries) barely count as literate. Perhaps half[1]. Most of the rest aren't a ton better, and can't write worth a damn. I think "on average" understates the case: at least 90% of US adults would probably get better results by passing their writing through ChatGPT, than not. Most of that remaining 10% would also see improvements from ChatGPT, except for writing they've put a great deal of effort into. It's very good at cleaning up writing, and/or tuning it to be better-suited to some purpose than the original was. In fact, a large majority of readers in the US would probably benefit from passing what they read through ChatGPT and having it summarize it concisely, in very simple terms. The world makes a lot more sense, to those who grew up taking to the written word as naturally as a duck to water, when we appreciate that the world runs on the written word, yet most people are terrible readers and worse writers. [1] "According to a 2020 report by the U.S. Department of Education, 54% of adults in the United States have prose literacy below the 6th-grade level." And, nb, if you're a good reader and you're thinking "6th grade level, that's not so bad", your idea of what a "6th-grade level" looks like, is probably skewed way higher than what they mean. https://en.wikipedia.org/wiki/Literacy_in_the_United_States https://en.wikipedia.org/wiki/Literacy_in_the_United_States

yyyk 3y ago
>at least 90% of US adults would probably get better results by passing their writing through ChatGPT, than not.... In fact, a large majority of readers in the US would probably benefit from passing what they read through ChatGPT Bullet list -> AI text generator -> AI text summarizer -> Bullet list Perhaps OpenAI can optimize the process by storing the original prompt and just giving it back at the last phase.
stOneskull 3y ago
> 54% of adults in the United States have prose literacy below the 6th-grade level. this is quite saddening.
- bamfly 3y ago

burnished 3y ago

I havent seen a compelling example of that, to the point that I think it has associated 'intelligence' with verbosity. I'd expect writing samples to get shorter, not longer, when being improved.