So spot on. As a blogger of 20 years with nearly 5500 posts under my belt, and one who has been using GPT-4 in earnest for months with my API plugged into several available tools (mostly Google Docs and the WordPress editor), there is so much more to do than have it just write stuff for you or to ask it things.
It just "juices me up," for lack of a better description. I principally use just "expand" and "rewrite" prompts. So, for instance, I'll sloppily draft a paragraph. Hit expand, and I get back 3-5 paragraphs nicely covering all the points.
Then I'll rewrite each of those, and then have GPT rewrite what I rewrote. At any point, I can expand a paragraph even further. Sometimes, just an off-the-top-of-head sentence turns into an entire section of a post.
That's a rough process that varies widely depending on what I'm doing...and I can stop any time and publish whatever the last iteration is.
It's essentially my writing partner. The key for me is not being lazy. I use it to do more, better. So for those who think AI is going to render that sort of writing of less value, I think the opposite. It's a better instrument, like going from an acoustic to electric guitar. Both still require the musician, but the latter allows for a great and "louder" range of output to a bigger and wider audience.
Happy Teachers' Day, Ethan! This isn't related to the above blog post, but you're probably one of the best teachers I've never met 🙏
For context: In India, we celebrate Teachers' Day on September 5th, and I make it a point to write to my teachers. You've been an incredible guide through your writings and videos.
Learning about AI can be overwhelming, especially with the flood of content on the internet. However, your work has given me direction and has actually helped me both think better and improve my work. Thank you for being a terrific educator and for devoting time to spread knowledge. Hat tip 🎩
Sometimes I take these types of checking tasks and turn them into dedicated bots in Poe based on either GPT-4 or Claude-2. So for example, I made a "no-passive-voice" bot based on your prompt here. Now I can just open it up and paste the text, and it does everything else.
This makes me think there will apps powered by LLMs with built-in prompts to accomplish specific goals like to improve writing. Or treat mood disorders.
Spot on! I hated creative writing because I lacked confidence in my writing, until GPT-4.
Now I feel like I have this super-weirdo creative writing partner to bounce ideas, expand, rewrite, and review all my weird ideas. The best part, it’s a partner without any ego or incentive biases.
I love writing now. And I’m excited to see the next thing AI will make me fall for next.
I am a frequent contributor to a tech trade publication. In the past, I would always have a reader or two help to find the rough spots. Now, however, AI has become my editing partner; not for grammar but to ensure the content is topical and cogent in addition to being well written. The responses I get are mostly spot-on, and this both saves time and improves the quality of my output. I cannot even imagine how this will change again when Gemini (5x performance gain?!) and other next-generation LLMs arrive.
Just once more, I'd like to pitch the idea of you using the current best version of AI Speech generation in order to increase accessibility for your articles. They are really helpful for thinking how to best utilise AI, and for a few people, that advice is even more valuable if they naturally struggle with text.
Here is a conversion of this article, utilising two voices, one for the main text and one for "AI".
This is an incredibly valuable piece of writing! I thank you for your contribution, and your conceptualization of AI and its utility is especially, encouraging. Of everything you wrote here my favorite is the following quote:
“These early days of AI are already absurd. Humans, walking and talking bags of water and trace chemicals that we are, have managed to convince well-organized sand to pretend to think like us. We don’t know what will happen next, and what dangers and opportunities the next generations of AI will bring. But we do need to realize that trying to convince ourselves that AI is normal software will not protect us from disruption. Instead, it may make it harder for us to see what is coming.”
I love that your explanations teach us how to think about making generative AIs much more useful! And, you very generously share useful examples proving your point. Thank you.
Is there maybe a danger that AI will give us links between ideas that aren't really there? Like the example of comparing a thesaurus to medieval craft guilds of 14th-century Europe? It makes a good comparison, but it's not something that you'd come to logically.
I wonder if AI is going to make radical ideas even more radical and give us more answers than we need, making everything more ambiguous.
So spot on. As a blogger of 20 years with nearly 5500 posts under my belt, and one who has been using GPT-4 in earnest for months with my API plugged into several available tools (mostly Google Docs and the WordPress editor), there is so much more to do than have it just write stuff for you or to ask it things.
It just "juices me up," for lack of a better description. I principally use just "expand" and "rewrite" prompts. So, for instance, I'll sloppily draft a paragraph. Hit expand, and I get back 3-5 paragraphs nicely covering all the points.
Then I'll rewrite each of those, and then have GPT rewrite what I rewrote. At any point, I can expand a paragraph even further. Sometimes, just an off-the-top-of-head sentence turns into an entire section of a post.
That's a rough process that varies widely depending on what I'm doing...and I can stop any time and publish whatever the last iteration is.
It's essentially my writing partner. The key for me is not being lazy. I use it to do more, better. So for those who think AI is going to render that sort of writing of less value, I think the opposite. It's a better instrument, like going from an acoustic to electric guitar. Both still require the musician, but the latter allows for a great and "louder" range of output to a bigger and wider audience.
I like the characterization "AI is like Google that lies" :)
although I prefer "it is like an overeager undergraduate
research assistant".
Happy Teachers' Day, Ethan! This isn't related to the above blog post, but you're probably one of the best teachers I've never met 🙏
For context: In India, we celebrate Teachers' Day on September 5th, and I make it a point to write to my teachers. You've been an incredible guide through your writings and videos.
Learning about AI can be overwhelming, especially with the flood of content on the internet. However, your work has given me direction and has actually helped me both think better and improve my work. Thank you for being a terrific educator and for devoting time to spread knowledge. Hat tip 🎩
Interesting take on the perspective on using AI like ChatGPT to expand our horizons instead of changing everything we have done before it.
I am currently practicing the usage of Chat GPT in the past few weeks and found this free open source website
https://learnprompting.org/
very useful.
Maybe some of your also find it helpful
An excellent and useful post as always, thank you!
Sometimes I take these types of checking tasks and turn them into dedicated bots in Poe based on either GPT-4 or Claude-2. So for example, I made a "no-passive-voice" bot based on your prompt here. Now I can just open it up and paste the text, and it does everything else.
This makes me think there will apps powered by LLMs with built-in prompts to accomplish specific goals like to improve writing. Or treat mood disorders.
Flipping this was great. I like the idea of ai as a writing partner than can not replace the human creativity, but sharpen it.
Spot on! I hated creative writing because I lacked confidence in my writing, until GPT-4.
Now I feel like I have this super-weirdo creative writing partner to bounce ideas, expand, rewrite, and review all my weird ideas. The best part, it’s a partner without any ego or incentive biases.
I love writing now. And I’m excited to see the next thing AI will make me fall for next.
I am a frequent contributor to a tech trade publication. In the past, I would always have a reader or two help to find the rough spots. Now, however, AI has become my editing partner; not for grammar but to ensure the content is topical and cogent in addition to being well written. The responses I get are mostly spot-on, and this both saves time and improves the quality of my output. I cannot even imagine how this will change again when Gemini (5x performance gain?!) and other next-generation LLMs arrive.
Creative people will be more creative with AI. Productive people more productive. Lazy people will use it lazily.
Just once more, I'd like to pitch the idea of you using the current best version of AI Speech generation in order to increase accessibility for your articles. They are really helpful for thinking how to best utilise AI, and for a few people, that advice is even more valuable if they naturally struggle with text.
Here is a conversion of this article, utilising two voices, one for the main text and one for "AI".
https://youtu.be/sqHGCDzNuBI
This is an incredibly valuable piece of writing! I thank you for your contribution, and your conceptualization of AI and its utility is especially, encouraging. Of everything you wrote here my favorite is the following quote:
“These early days of AI are already absurd. Humans, walking and talking bags of water and trace chemicals that we are, have managed to convince well-organized sand to pretend to think like us. We don’t know what will happen next, and what dangers and opportunities the next generations of AI will bring. But we do need to realize that trying to convince ourselves that AI is normal software will not protect us from disruption. Instead, it may make it harder for us to see what is coming.”
I love that your explanations teach us how to think about making generative AIs much more useful! And, you very generously share useful examples proving your point. Thank you.
Is there maybe a danger that AI will give us links between ideas that aren't really there? Like the example of comparing a thesaurus to medieval craft guilds of 14th-century Europe? It makes a good comparison, but it's not something that you'd come to logically.
I wonder if AI is going to make radical ideas even more radical and give us more answers than we need, making everything more ambiguous.
I haz more words now. Thanks AI and your helpful care assistant Ethan…