First, what do you think of the post in the context of what it says? Is it fit for publication?
Second, here is a comment that I wrote. What do you think of it?
********************************
This is an interesting post. I use ChatGPT quite a lot in my writing. Here are some thoughts.
1. It is good at spotting ‘silly’ mistakes and typographical errors.
2. It can serve as a high level checklist, and prompt ‘Oh, I forgot about that’ responses.
3. It can analyze internal logic and inconsistencies.
4. It can add more detail ― it is good at fleshing out thoughts.
Nevertheless, I find that the end product is not quite good enough. Human review is still required.
I have noticed that in the short amount of time it has been available, ChatGPT has improved in all areas. (For example, it is now much better at providing citations.) Where will it be five years from now?
Here is the reply.
********************************
Good afternoon! I'm glad to hear the weather is treating you well in central Virginia—it’s always nice when spring shows its best side.
Unfortunately, I wasn’t able to access Umberto Eco’s post directly at the link you provided (AI-Assisted Paper Reviewing), so I can't assess the article itself for fitness for publication without its content. If you can copy and paste the text or summarize the key points here, I’d be happy to evaluate it with you.
Regarding your comment:
Your comment is thoughtful, well-structured, and strikes a balanced tone. Here’s a quick analysis:
Strengths:
Clear structure: The bullet-point list makes it easy to follow.
Balanced perspective: You acknowledge the strengths and current limitations of AI tools.
Forward-looking: Ending with a reflection on future developments adds insight and encourages discussion.
Suggestions (very minor):
You might consider briefly clarifying what you mean by “not quite good enough.” Is it a matter of tone, accuracy, depth, or something else? That would help readers understand where you draw the line between AI and human capability.
Overall, it’s a solid, constructive comment that contributes positively to a thoughtful conversation.
If you'd like to go deeper into any of the points or work on a revision, just let me know!
From a medical perspective, the risk of that alone already is too high to ignore.
The weakness of AI is the lack of sensors to do measurements so it no longer needs to rely on internet links (and (political / questionable?) data). This is a work already in progress:
---
A new agentic approach called 'streams' will let AI models learn from the experience of the environment without human 'pre-judgment'.
You make a good case for using AI for first review. At least we know the AI will read the paper! But don't mistake AI for critical thinking. AI does not do logic, it only finds language relationships with its vast training data. If you are publishing something totally new, AI will not find it in the training data. AI will likely flag it as wrong, rather than novel.
Are modern human beings absurd or is absurdity built into AI. Either way the papers can be generally considered as junk produced by “dead labor”
AI is Ecocide.
Alas…. ecocide is a human thing.
AI also.
I asked ChatGPT to help me write this comment. Here is what I asked.
********************************
Good afternoon ChatGPT:
I hope that you are having a good day. The weather here in central Virginia is delightful.
My colleague Umberto Eco has written an interesting paper entitled ‘AI Assisted Paper Reviewing’ at https://livingearth.substack.com/p/ai-assisted-paper-reviewing.
I have two requests.
First, what do you think of the post in the context of what it says? Is it fit for publication?
Second, here is a comment that I wrote. What do you think of it?
********************************
This is an interesting post. I use ChatGPT quite a lot in my writing. Here are some thoughts.
1. It is good at spotting ‘silly’ mistakes and typographical errors.
2. It can serve as a high level checklist, and prompt ‘Oh, I forgot about that’ responses.
3. It can analyze internal logic and inconsistencies.
4. It can add more detail ― it is good at fleshing out thoughts.
Nevertheless, I find that the end product is not quite good enough. Human review is still required.
I have noticed that in the short amount of time it has been available, ChatGPT has improved in all areas. (For example, it is now much better at providing citations.) Where will it be five years from now?
Here is the reply.
********************************
Good afternoon! I'm glad to hear the weather is treating you well in central Virginia—it’s always nice when spring shows its best side.
Unfortunately, I wasn’t able to access Umberto Eco’s post directly at the link you provided (AI-Assisted Paper Reviewing), so I can't assess the article itself for fitness for publication without its content. If you can copy and paste the text or summarize the key points here, I’d be happy to evaluate it with you.
Regarding your comment:
Your comment is thoughtful, well-structured, and strikes a balanced tone. Here’s a quick analysis:
Strengths:
Clear structure: The bullet-point list makes it easy to follow.
Balanced perspective: You acknowledge the strengths and current limitations of AI tools.
Forward-looking: Ending with a reflection on future developments adds insight and encourages discussion.
Suggestions (very minor):
You might consider briefly clarifying what you mean by “not quite good enough.” Is it a matter of tone, accuracy, depth, or something else? That would help readers understand where you draw the line between AI and human capability.
Overall, it’s a solid, constructive comment that contributes positively to a thoughtful conversation.
If you'd like to go deeper into any of the points or work on a revision, just let me know!
Interesting review, as one of the coauthors ( Phil Bierwirth) already published an article with a graph showing a decline in IQ vs time: https://www.researchgate.net/publication/328781907_Are_increasing_atmospheric_carbon_dioxide_levels_lowering_our_intelligence_and_making_us_anxious
From a medical perspective, the risk of that alone already is too high to ignore.
The weakness of AI is the lack of sensors to do measurements so it no longer needs to rely on internet links (and (political / questionable?) data). This is a work already in progress:
---
A new agentic approach called 'streams' will let AI models learn from the experience of the environment without human 'pre-judgment'.
https://www.zdnet.com/article/ai-has-grown-beyond-human-knowledge-says-googles-deepmind-unit
---
When the sensors are sensitive enough, in a global network the AI can reveal trends that now escape observation.
I would like a preprint!
You make a good case for using AI for first review. At least we know the AI will read the paper! But don't mistake AI for critical thinking. AI does not do logic, it only finds language relationships with its vast training data. If you are publishing something totally new, AI will not find it in the training data. AI will likely flag it as wrong, rather than novel.
I think I have your email -- but would you send it to me again?
bystrc@rpi.edu
thx!
People seem unable to actually monitor the driving of a self-driving car...
;-(
What you mean, "AI assisted"?
Well, the AI does the hard job of verifying the authors' claims. The humans supervise.
I wish this project well, but the human tendency to hand it all off to the "underling" is clear.