<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>ChatGPT on</title><link>https://geekyschmidt.com/categories/chatgpt/</link><description>Recent content in ChatGPT on</description><image><title/><url>https://geekyschmidt.com/images/papermod-cover.png</url><link>https://geekyschmidt.com/images/papermod-cover.png</link></image><generator>Hugo</generator><language>en</language><copyright>Copyright ©2002-2026, Nicholas Schmidt; all rights reserved.</copyright><lastBuildDate>Thu, 21 Sep 2023 04:15:57 +0000</lastBuildDate><atom:link href="https://geekyschmidt.com/categories/chatgpt/index.xml" rel="self" type="application/rss+xml"/><item><title>ChatGPT and the Electric Bill Whisperer - How to Upload Your Bills and Watch Tariff Magic Happen</title><link>https://geekyschmidt.com/post/2023-09-21-chatgpt-energy-bills/</link><pubDate>Thu, 21 Sep 2023 04:15:57 +0000</pubDate><guid>https://geekyschmidt.com/post/2023-09-21-chatgpt-energy-bills/</guid><description>&lt;p&gt;I will do a little experiment and let ChatGPT also write this blog post. I have made minor edits to ensure it aligns with the screenshots, but this is not me!&lt;/p&gt;
&lt;h3 id="a-few-essential-items-before-we-dive-in"&gt;A few essential items before we dive in:&lt;/h3&gt;
&lt;ul&gt;
&lt;li&gt;I am using GPT4 with Data Analysis through a paid account&lt;/li&gt;
&lt;li&gt;I work in the energy space, so I knew basically what I was trying to do with the prompts&lt;/li&gt;
&lt;li&gt;The Consumers website could have done this analysis in precisely 3 clicks through their portal, so I saved little time&lt;/li&gt;
&lt;li&gt;Yes, we use a ton of electricity! We have a large geothermal system, 22kw solar array, and 26kw of batteries in the basement. We also have 2 Electric Vehicles and a manufacutring shop on the land.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="chatgpt-shortcomings"&gt;ChatGPT Shortcomings:&lt;/h3&gt;
&lt;ol&gt;
&lt;li&gt;There is no confidence rating here. You ask the model; it gives answers. Are those answers correct? Sometimes. There is a fallacy that you upload and ask away. In another exercise, I needed to go back and validate the PDF breakout, and the table read needed to be corrected. &lt;em&gt;&lt;strong&gt;Verify everything.&lt;/strong&gt;&lt;/em&gt;&lt;/li&gt;
&lt;li&gt;For a site-by-site analysis, this works great, but scaling this would require API integration. Back to my first point here, you would trust that automation wholeheartedly.&lt;/li&gt;
&lt;li&gt;As technology advances, the ability to quickly ingest and distil complex tariff documents will be a net positive to the energy industry. We are close, but not there just yet.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="steps-i-used-to-complete-this-exercise"&gt;Steps I used to complete this exercise:&lt;/h3&gt;
&lt;ol&gt;
&lt;li&gt;I went to GreenButton Connect for Consumers and downloaded my interval and billing data. I tried using some lower-level interval data from my parallel metering, but it was too small of an interval to process correctly.&lt;/li&gt;
&lt;li&gt;Downloaded the latest tariff book for all Consumers Energy rates.&lt;/li&gt;
&lt;li&gt;Logged into ChatGPT, selected my GPT4 analysis mode and provided the prompt - &lt;em&gt;How does my utility bill compare to other people in Michigan?&lt;/em&gt; before starting to have it evaluate the usage against the tariff specifically.&lt;/li&gt;
&lt;li&gt;Reviewed the Python code and processing&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;You can see the entire chat minus the actual code here: &lt;a href="https://chat.openai.com/share/fe77f5e1-b65f-4c52-a3f7-ef5fe2e7acb0"&gt;ChatGPT Chat Prompts&lt;/a&gt;&lt;/p&gt;</description></item></channel></rss>