Adding an Advice From OpenAI Field to my Journal
I'm a cyborg who is exploring the value of OpenAI, utilizing AI-assisted thought processes and hotkeys to turn the AI on and off. I'm testing a function to give advice from the OpenAI API and am using it to summarize blog posts into meta descriptions, headlines, and keywords, as well as suggest life improvements.
Exploring the Value of OpenAI: My Journey as a Cyborg with AI-Assisted Thought Processes
By Michael Levin
Thursday, April 27, 2023
I lost time lately on turning myself into a cyborg. My thought process is AI-assisted now by default. Yes, I keep Copilot on by default again now in journaling, but I have a F6 hotkey that will turn it off and an F7 hotkey that will turn it back on. F6 on my laptop keyboard is “dimmer” and F7 is “brighter”, haha! I thought that appropriate.
Okay, given that OpenAI is now a co-author on my journal, I’m going to have to explore what more value I can get from it than just summarizing the posts into meta descriptions, headlines and keywords. I should have it read each blog post with a thoughtful eye and suggest life improvements. I must design a new prompt and have the response added to the YAML front matter of the post just like with descriptions, headlines and keywords.
You are my work advisor and life-coach. Read what I have written in the context of everything else at https://mikelev.in/blog/ and tell me what I should do next. My goal is financial independence and achieving my ikigai.
A question is whether I should have it read as much as it can directly from the blog post, but not the whole thing because I don’t want to pay for that much token usage again, or whether it can do a sufficient job reading the already rolled-up summaries? I think I’ll start with the summaries and see how it goes.
I’m going to have to start a new journal post for this, because when I add a new field to the YAML front matter, I don’t want it reprocessing everything retroactively. Okay, I wrote this function:
@retry(Exception, delay=1, backoff=2, max_delay=60)
def give_advice(data):
"""Write a meta description for a post."""
response = openai.Completion.create(
engine=ENGINE,
prompt=(
f"You are my work advisor and life-coach. "
"Read what I have written in the context of everything else at "
"https://mikelev.in/blog/ "
"and tell me what I should do next:\n{data}\n\n"
"My goal is financial independence and achieving my ikigai. "
"\nAdvice:\n\n"
),
temperature=0.5,
max_tokens=100,
n=1,
stop=None,
)
advice = response.choices[0].text.strip()
return advice
It will be interesting this being the first post for which I’m asking OpenAI API (GPT-4, presumably) to give me advice with an automated script. I have done it through the chat.openai.com site before, but this is life-integration now.
My existing code should create the new advice.db if it doesn’t exist (which it doesn’t) on the first run. Let’s test…