I think that's the cultural differences driving you to think about the right way. We all have our own biases and there is a bias interaction in prompts when we provide our subjective take on expanding our ideas in the right away from the conceptual framework that is the language.
I think using LLMs to help review your work is huge. I don't like publishing anything without having someone read it to make sure I didn't miss anything obvious; usually an LLM is way easier to find (and takes far less time) than a human.
When it comes to tools for writing, there are several options that I recommend trying out. One of them is ChatGPT, which is useful for generating conversational responses, answering questions, and providing information on a wide range of topics. Additionally, there are other tools that can be helpful in various aspects of writing. Grammarly is a popular choice for checking grammar, spelling, and style. It offers real-time suggestions and corrections, enhancing the clarity and correctness of your writing.
This is the current global English linguistic infrastructure that rules the world currently. French and Spanish are probably up there as well. I tend to use ChatGPT to put my thoughts into a paragraph and see what connects these ideas as I see the similarity and connection through my creative ideation, and then research more about the connections that AI lays out in front of me. Expanding and summarising tells me the priority or different words, that they have different subjective values to communities in the world that make this tech.
One such concern associated with technologizing language models is reinforcement bias. Since they learn from existing data, including human-generated text from the internet, they can inadvertently amplify or reinforce existing biases present in the data. Language, being deeply intertwined with culture, dialects, and linguistic variations, presents additional challenges. Another challenge is the potential for language models to be culturally inclined towards specific audiences or cater more to certain linguistic norms. This can lead to a lack of representation and understanding of different cultural perspectives, which can be problematic in a global and diverse society.
Greed, driven by desires for wealth and power, can influence AI decision-making. Attention bias in algorithms and platforms prioritizes content that captures attention, potentially reinforcing existing beliefs and limiting exposure to diverse perspectives. Conscious engagement, which involves thoughtful participation and decision-making, can be disrupted by AI systems optimized for metrics like engagement and user retention. It is crucial to recognize that AI lacks consciousness and intentionality. Responsible AI development requires aligning systems with human values, promoting transparency, accountability, and incorporating diverse perspectives. Fostering conscious engagement and critical thinking skills helps individuals navigate AI's influence. By addressing these concerns, we can harness AI's benefits while minimizing negative consequences.
Come offer your thoughts on AI assistance in digital content creation moving forward.
The idea is to be through in my deepness. Not to leave you at the deep end but cover the deep end and try to explain more about how I can relate it to the simple. I'm slowly getting better and figuring out a formula to introduce more human insights while the language and writing is handled by the AI. The smaller responses are more likely me personally but sometimes my discussion with AI can get deep, so I try to bridge the gap to simplicity by connecting it back to my personal experiences. That's how I've tried to improve my content creation and automated my process flow. I'd appreciate your support and feedback so I can even create a more standard template etc.
Finally, it wasn't just a discussion for the tech writers but a call for the larger substack community to help with this discussion. I want people to have deeper discussions as automation complexifies the value exchange and we contribute to various enablers in society knowingly or unknowingly. Just as we borrow concepts from the past, we have to adapt to the contexts of today and Discuss the cultural heritage in a responsible and respectful manner. The status quo always feels oppressive but also it's imperative to just understand their dominance and grow beyond it. That's how I see diverse cultures producing strong healthy Individuals who can be global citizens
I don't particularly like content written by AI, I seem to have developed a nose for detecting it.
However, I use it to review my writing. Even that usage is getting reduced as I write more and more. GPT-3.5 is pretty much useless for this work, it will find issues which are not there. GPT-4 does a much better job but it can be unreliable at times. I do one round of review with it, fix some of the issues it finds, and then get another round of proof-reading done.
Sometimes I would use it to explore potential ideas for my next topic, however I prefer not to ask for an outline so as not to color my vision of how it should be written. I have found that it is not very imaginative or creative in suggesting outlines, and you will find numerous other articles with similar structure already out there. So I prefer to explore my own angle as I write on that topic.
I also use it for suggesting titles for my articles, and even that I am cutting down. Its title ideas are also very predictable (again GPT-4 is slightly better here). For example it suggests titles like "Unleashing the power of <xyz>". In my experience, titles for emails should be such that the subscriber will be curious to open it, while for online audience should be something very clear and meaningful so that they will want to open the link if they see it on Google.
I did ask ChatGPT to proof-read it, and it rewrote it in a way that reads very robotic, so I kept my original as it is. Here is what it had rewritten:
---------------
I don't particularly enjoy content written by AI; it seems I have developed a knack for detecting it. However, I do utilize it for reviewing my own writing, although I find myself relying on it less and less. GPT-3.5 is largely ineffective for this purpose as it often identifies issues that aren't actually present. On the other hand, GPT-4 performs better but can still be unreliable. Typically, I conduct one round of review with GPT-4, address some of the issues it highlights, and then seek another round of proofreading.
Occasionally, I use AI to explore potential ideas for my next topic. However, I prefer not to request an outline in order to avoid influencing my own perception of how it should be written. I've discovered that AI isn't particularly imaginative or creative when it comes to suggesting outlines, and you'll likely find numerous other articles with similar structures already available. Therefore, I prefer to establish my own unique angle as I write.
Furthermore, I rely on AI for generating title suggestions for my articles, but even in that aspect, I'm reducing my dependence on it. AI's title ideas are often predictable (although GPT-4 is slightly better in this regard). For instance, it tends to propose titles such as "Unleashing the power of <xyz>," whereas titles for emails should pique subscribers' curiosity and encourage them to open the email. On the other hand, titles for online audiences should be clear, meaningful, and compelling enough to entice them to click on the link if they encounter it on a search engine like Google.
Haha it uses a lot of Occasionally, Furthermore, Additionally, Utilize... and other kinda verbose wording. Yeah my experience is the same, for formal writing you can kind of let it free but for more personal writing it quickly devolves into something dry.
Well mostly I use it to talk through the most optimal way to approach a particular topic, for instance, when I was working on the Brain Inspired Approach to AI essay I didn't know how to approach it as a single body due to the fact that the research I did on the brain was really comprehensive, so GPT pointed out that rather than attempting to link the Brain and AI systems I could use the History of how the BraiN Inspired came to be as a link. I use GPT to refine my approach to content and of course sources of research.
Going on a tangent, do you think the creativity of Text Generators is limited to the skill of the prompter or the due to a certain narrow view of data available?
It is very useful to research key points to help confirm I am thinking about things the right way.
Which tool(s) do you use? Directly ChatGPT or something task-specific?
Just ChatGPT
Got it. So I imagine you have to copy and paste and move back and forth between the chat UI and whatever you're writing?
I rarely copy stuff. Sometimes. Usually just using it for ideas, then I write in my own voice.
I think that's the cultural differences driving you to think about the right way. We all have our own biases and there is a bias interaction in prompts when we provide our subjective take on expanding our ideas in the right away from the conceptual framework that is the language.
Oh wow, I literally just wrote about this:
https://goatfury.substack.com/p/can-ai-be-funny
I think using LLMs to help review your work is huge. I don't like publishing anything without having someone read it to make sure I didn't miss anything obvious; usually an LLM is way easier to find (and takes far less time) than a human.
When it comes to tools for writing, there are several options that I recommend trying out. One of them is ChatGPT, which is useful for generating conversational responses, answering questions, and providing information on a wide range of topics. Additionally, there are other tools that can be helpful in various aspects of writing. Grammarly is a popular choice for checking grammar, spelling, and style. It offers real-time suggestions and corrections, enhancing the clarity and correctness of your writing.
This is the current global English linguistic infrastructure that rules the world currently. French and Spanish are probably up there as well. I tend to use ChatGPT to put my thoughts into a paragraph and see what connects these ideas as I see the similarity and connection through my creative ideation, and then research more about the connections that AI lays out in front of me. Expanding and summarising tells me the priority or different words, that they have different subjective values to communities in the world that make this tech.
One such concern associated with technologizing language models is reinforcement bias. Since they learn from existing data, including human-generated text from the internet, they can inadvertently amplify or reinforce existing biases present in the data. Language, being deeply intertwined with culture, dialects, and linguistic variations, presents additional challenges. Another challenge is the potential for language models to be culturally inclined towards specific audiences or cater more to certain linguistic norms. This can lead to a lack of representation and understanding of different cultural perspectives, which can be problematic in a global and diverse society.
Greed, driven by desires for wealth and power, can influence AI decision-making. Attention bias in algorithms and platforms prioritizes content that captures attention, potentially reinforcing existing beliefs and limiting exposure to diverse perspectives. Conscious engagement, which involves thoughtful participation and decision-making, can be disrupted by AI systems optimized for metrics like engagement and user retention. It is crucial to recognize that AI lacks consciousness and intentionality. Responsible AI development requires aligning systems with human values, promoting transparency, accountability, and incorporating diverse perspectives. Fostering conscious engagement and critical thinking skills helps individuals navigate AI's influence. By addressing these concerns, we can harness AI's benefits while minimizing negative consequences.
Come offer your thoughts on AI assistance in digital content creation moving forward.
Deep and thorough! Can I ask how much if this is generated vs directly written by yourself? Honest question.
80:20 pareto effort. I use soem sentences and then I write or edit the rest.
Got it :)
The idea is to be through in my deepness. Not to leave you at the deep end but cover the deep end and try to explain more about how I can relate it to the simple. I'm slowly getting better and figuring out a formula to introduce more human insights while the language and writing is handled by the AI. The smaller responses are more likely me personally but sometimes my discussion with AI can get deep, so I try to bridge the gap to simplicity by connecting it back to my personal experiences. That's how I've tried to improve my content creation and automated my process flow. I'd appreciate your support and feedback so I can even create a more standard template etc.
Finally, it wasn't just a discussion for the tech writers but a call for the larger substack community to help with this discussion. I want people to have deeper discussions as automation complexifies the value exchange and we contribute to various enablers in society knowingly or unknowingly. Just as we borrow concepts from the past, we have to adapt to the contexts of today and Discuss the cultural heritage in a responsible and respectful manner. The status quo always feels oppressive but also it's imperative to just understand their dominance and grow beyond it. That's how I see diverse cultures producing strong healthy Individuals who can be global citizens
I don't particularly like content written by AI, I seem to have developed a nose for detecting it.
However, I use it to review my writing. Even that usage is getting reduced as I write more and more. GPT-3.5 is pretty much useless for this work, it will find issues which are not there. GPT-4 does a much better job but it can be unreliable at times. I do one round of review with it, fix some of the issues it finds, and then get another round of proof-reading done.
Sometimes I would use it to explore potential ideas for my next topic, however I prefer not to ask for an outline so as not to color my vision of how it should be written. I have found that it is not very imaginative or creative in suggesting outlines, and you will find numerous other articles with similar structure already out there. So I prefer to explore my own angle as I write on that topic.
I also use it for suggesting titles for my articles, and even that I am cutting down. Its title ideas are also very predictable (again GPT-4 is slightly better here). For example it suggests titles like "Unleashing the power of <xyz>". In my experience, titles for emails should be such that the subscriber will be curious to open it, while for online audience should be something very clear and meaningful so that they will want to open the link if they see it on Google.
Interesting! Can I ask how much of this very reply is AI generated? Or at least revised? Just an honest question to get some statistics.
I did ask ChatGPT to proof-read it, and it rewrote it in a way that reads very robotic, so I kept my original as it is. Here is what it had rewritten:
---------------
I don't particularly enjoy content written by AI; it seems I have developed a knack for detecting it. However, I do utilize it for reviewing my own writing, although I find myself relying on it less and less. GPT-3.5 is largely ineffective for this purpose as it often identifies issues that aren't actually present. On the other hand, GPT-4 performs better but can still be unreliable. Typically, I conduct one round of review with GPT-4, address some of the issues it highlights, and then seek another round of proofreading.
Occasionally, I use AI to explore potential ideas for my next topic. However, I prefer not to request an outline in order to avoid influencing my own perception of how it should be written. I've discovered that AI isn't particularly imaginative or creative when it comes to suggesting outlines, and you'll likely find numerous other articles with similar structures already available. Therefore, I prefer to establish my own unique angle as I write.
Furthermore, I rely on AI for generating title suggestions for my articles, but even in that aspect, I'm reducing my dependence on it. AI's title ideas are often predictable (although GPT-4 is slightly better in this regard). For instance, it tends to propose titles such as "Unleashing the power of <xyz>," whereas titles for emails should pique subscribers' curiosity and encourage them to open the email. On the other hand, titles for online audiences should be clear, meaningful, and compelling enough to entice them to click on the link if they encounter it on a search engine like Google.
-------------------
Haha it uses a lot of Occasionally, Furthermore, Additionally, Utilize... and other kinda verbose wording. Yeah my experience is the same, for formal writing you can kind of let it free but for more personal writing it quickly devolves into something dry.
Yes, exactly.
I enjoy the writing too much to farm it out to GPT, and it would be really difficult to get it to write in my style, anyway.
I do use GPT for technical reviews of the article as a whole (not great) and of smaller sections (better, with the right prompting)
I also use it to generate ideas, or more specifically to expand on some ideas I may already have.
Well mostly I use it to talk through the most optimal way to approach a particular topic, for instance, when I was working on the Brain Inspired Approach to AI essay I didn't know how to approach it as a single body due to the fact that the research I did on the brain was really comprehensive, so GPT pointed out that rather than attempting to link the Brain and AI systems I could use the History of how the BraiN Inspired came to be as a link. I use GPT to refine my approach to content and of course sources of research.
Going on a tangent, do you think the creativity of Text Generators is limited to the skill of the prompter or the due to a certain narrow view of data available?