Tag: AI

Posted on

Attempt to Use AI to Write My Biography

I doubt if there are many people interested my biography. I’m doing this as an experiment to see what AI will write, just as I tried to use SEO to get a picture of my Alfetta Sprint Veloce to come up in a Google image search.

I got the idea from Tyler Cowen – to have AI write my biography. I thought that Substack would be a good place for it because I could write biographical things on it exclusively. I thought that AI crawlers would eventually learn about me if I used the slug, “mrlesliewong” on everything.

I started by doing things to optimize my Substack for search. I put my Google Tag Manager ID into the analytics settings on my Substack settings. I added mrlesliewong.substack.com as a new property in Google Search Console using URL Prefix Verification. Substack even has a post, A guide to SEO on Substack. But then I started reading about the difficulty with SEO with Substack, (e.g. the subdomain structure. this thread on Reddit, etc)

I asked ChatGPT if I would I have a better chance of my writing being crawled by AI models for training data if I posted on my own blog, https://www.lesliewong.us/ instead of Substack. The answer:

“Yes, posting on your own blog at https://www.lesliewong.us/ (assuming it’s publicly accessible and indexed by search engines) would likely give you a better chance of having it included in AI training data than posting only on Substack. Here’s why:

Why Your Own Blog is Better for AI Training Data:
Search Engine Indexing: Since you’ve set up lesliewong.us in Google Search Console, Google is aware of your site. This increases the likelihood of your content being crawled, stored, and potentially included in datasets used for AI training.
Web Scraping for AI Training: Many AI models are trained on publicly available websites that search engines index. Your blog, being independent, could be part of these datasets, whereas Substack content might not be as widely scraped.
Full Control Over Content: On your own blog, you control how long the content remains available, ensuring it persists over time for potential inclusion in AI models.
Customization & Metadata: You can add structured data (schema markup) that describes your content explicitly, making it more machine-readable.
Wider Reach Through Backlinks: If you link to your blog from social media, other blogs, or forums, search engines see it as more relevant.“

Because I bought a Pixel 9 Pro, Google gave me a Google One AI Premium subscription for a year. I asked Gemini the same question:

Yes, absolutely. Posting that content on your own blog, https://www.lesliewong.us/, would significantly increase its chances of being found and indexed by search engines, including being potentially used by AI models for training data. Here’s why:

Advantages of Your Own Blog:

Full Control:
You have complete control over your blog’s structure, content, and SEO. This allows you to optimize it for search engines.
You can implement proper heading tags (H1, H2, etc.), meta descriptions, and alt text for images, all of which are crucial for SEO.
Domain Authority:
Building a strong domain authority for lesliewong.us through consistent, high-quality content will improve your search engine rankings over time.
Google places a higher trust value on established domains.
Sitemap Submission:
You can create and submit a sitemap to Google Search Console, which helps Google crawl and index your content more efficiently. Substack sitemaps are problematic.
Structured Data:
You can implement structured data (schema markup) to provide search engines with more context about your content, making it easier for them to understand and display it in search results.
Long-Term Visibility:
Content on your own blog is more likely to remain accessible and visible over the long term, compared to content on a platform like Substack, where changes in platform policies or algorithms could affect visibility.
AI Model Training:
AI models are trained on large datasets of text and code from the internet. Content on your own website is more likely to be included in these datasets than content on a platform with certain restrictions.
Having your own domain allows your content to be seen as more authorative (sic).
Substack Limitations:

Platform Dependency:
You’re reliant on Substack’s algorithms and policies, which can change at any time.
SEO customization is limited.
Limited Control:
You have less control over the technical aspects of SEO, such as sitemap submission and structured data.

Recommendation:

Post your genealogical information on your https://www.lesliewong.us/ blog.
Optimize the content for SEO, including using relevant keywords, heading tags, and alt text for images.
Ensure your blog is mobile-friendly.
Promote your blog content through social media and other channels.
Continue to use substack, but understand that your personal web page will provide better long term search results.
By doing so, you’ll maximize the chances of your content being found by search engines and potentially used by AI models.

I also modified the robots.txt file on lesliewong.us to allow AI crawlers”

#Allow AI Crawlers

User-agent: GPTBot
Allow: /
User-agent: ChatGPT-User
Allow: /
User-agent: Googlebot
Allow: /
User-agent: Google-Extended
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: Amazonbot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: Bingbot
Allow: /

So instead of using Substack, I’m going to do that here, now. I’m also going to cross-post here, my two Substack posts.

Posted on

Using Gemini Advanced to Setup a Raspberry Pi Webcam

A month ago, a power outage scrambled my Raspberry Pi Zero 2 W running a webcam and it stopped working. I decided to try to restore it using the latest Raspberry Pi OS, Debian version: 12 (bookworm). I thought I’d try to use Artificial intelligence (AI) to help me set it up.

I’ve tried several generative artificial intelligence chatbots: Gemini Advanced, ChatGPT (3.5), Microsoft Copilot and Perplexity AI.  Google had offered a three month free trial of Gemini Advanced so I decided to try to use it to help me. I didn’t try GPT-4 or Copilot Professional, the paid chatbot products from OpenAI and Microsoft.

Raspberry Pi 4 webcam on the Larkspur, California bike path
Raspberry Pi 4 webcam using Raspberry Pi High Quality Camera aimed at the Larkspur Bike Path

Though I’ve been exposed to Linux, the Raspberry Pi was my first experience with having to write snippets of code (which I don’t know how to do) to get something to work. Sadly, my Raspberry Pi OS skills have not evolved as much as the Raspberry Pi hardware and I still search Google to find out how to do something (or look at my notes).

Twelve years ago, I managed to get a working webcam using a Logitech USB camera attached to a Model B using MJPG-streamer. I eventually started using Motion and the Raspberry PI High Quality Camera. The official Raspberry Pi site and Pi My Life up were two good places that got me started.

I ran into problems trying to use Bookworm, the camera and the Pi Zero 2 W itself (I broke the very fragile ribbon cable retain clip on the FPC connector). I then tried Bullseye and couldn’t get it to work either. I finally installed Buster, which was the last OS that worked and I still had problems getting the camera recognized. My troubles were compounded by the fact that I was trying to set it up headless, because, at the time, I didn’t have a mini HDMI to HDMI cable which the Pi Zero uses.

I decided to try to use a more robust Raspberry Pi 4 and Bookworm, (even though I had Motion running on the Zero 2 W with Buster had been running).

I installed Bookworm (64 bit) on the Pi 4 and started with Gemini. This was my first prompt:

I have a raspberry pi 4 running bookworm. I have a Raspberry Pi High Quality camera connected. When I run raspi-cam hello or libcamera-hello, I get no cameras available. In /boot/firmware/config.txt, I added the line dtoverlay=imx477

Gemini:

Absolutely, let’s troubleshoot this Raspberry Pi camera issue. Here’s a breakdown of potential causes and solutions:

  1. Physical Connections

Cable: Double-check the ribbon cable’s connections between the camera and the Pi board. Ensure it’s inserted fully and that the blue side faces the correct direction (toward the HDMI port on the Pi board).
Orientation: The metal contacts on the ribbon cable should face the HDMI port of your Raspberry Pi.
Damage: Inspect the cable and connectors for any signs of damage.

  1. Software Configuration

Legacy Support: Since you’re using Bookworm, it’s likely using the newer libcamera framework. If you had it running on an older Raspberry Pi OS version before, you may need to enable Legacy Camera support:

Run sudo raspi-config
Go to Interface Options
Select Legacy Camera
Enable it, then reboot.
dtoverlay: You mentioned adding dtoverlay=imx477. This overlay is specific to the IMX477 sensor used in the Raspberry Pi High Quality Camera. Ensure there are no typos in your /boot/firmware/config.txt file.

  1. Enabling the Camera Interface

raspi-config:
Run sudo raspi-config
Go to Interface Options
Select Camera
Ensure it’s enabled, then reboot.

  1. Firmware and Updates

Outdated Software: Make sure everything is up-to-date:
Bash
sudo apt update
sudo apt full-upgrade
Use code with caution.

  1. Hardware Issues

Camera Defect: If all else fails, there might be a hardware problem with the camera. If possible, try testing it on a different Raspberry Pi or try a different camera with your current Raspberry Pi.
Additional Troubleshooting

Permissions: Ensure your user has permissions to access the camera device (usually /dev/video0 ). You might try adding your user to the ‘video’ group:
Bash
sudo usermod -aG video
Use code with caution.
Conflict Check: Disable any other software that might be using the camera interface.
Let’s Get Specific:

Confirm Camera Model: Please double-check if you truly have the Raspberry Pi High Quality Camera (which uses the IMX477 sensor).
Error Messages: Provide any exact error messages you receive when running libcamera-hello or raspi-still.
Let me know the results of these checks, and we’ll narrow down the problem further!


Me: There is no Camera under Interface Options in raspi-config


Gemini:

You’re absolutely right! I apologize for that oversight. Here’s why the Camera option might be missing in raspi-config and how we can fix it.

Background:

Raspberry Pi OS has transitioned to the libcamera framework from the older raspicam system. This change streamlines camera access and provides more modern features. As a result, the legacy camera interface option has been removed from raspi-config in newer versions.

Solutions:

Confirming libcamera Support:

Run the command vcgencmd get_camera. You should see output like:
supported=1 detected=1
If detected=0, proceed with further troubleshooting.
Troubleshooting Steps:

Physical Connections: Recheck the ribbon cable connections and orientation as described in my previous response.
Updating Firmware: Sometimes firmware issues can cause problems. Use the following commands:
Bash
sudo apt update
sudo apt full-upgrade
sudo rpi-update
Use code with caution.
Reboot after these updates.
Checking the Device Tree: Make sure the camera overlay is enabled. If you haven’t already added it, edit /boot/firmware/config.txt and add this line:
dtoverlay=imx477
Reboot to apply.
Testing with libcamera:

Try using libcamera commands directly:

libcamera-still -t 5000 -o test.jpg (This should capture an image)
libcamera-hello (This should provide a preview from the camera)
Let’s Get More Info:

Could you specify the exact Raspberry Pi OS version you are using? You can find this by running cat /etc/os-release.
Share any error messages you specifically get from libcamera-still or libcamera-hello.
Let’s work together to get your camera up and running!

It pretty much went on like that for hours – I would do one thing Gemini recommended and it didn’t work. After more troubleshooting, there’d be a step forward, but then something else wouldn’t work.

Gemini always sounded upbeat, like we’d be able to make it work. I would upload my /etc/motion/motion.conf file, for example, Gemini would analyze the parameters, and give me suggestions.

“This output provides very useful clues!” “This is fantastic news!” ‘We’ll get this working smoothly!” … Gemini Advanced after I’d give it feedback

Sometimes, Gemini was helpful, but never to the point where I was able to achieve my initial goal of using a Pi 4 and the High Quality Camera with Bookworm.

This is a link to the Microsoft Word formatted conversation with Gemini Advanced. It’s 45,000+ words:

Conversation with Gemini Advanced to setup a Raspberry Pi WebcamDownload

I started searching for answers using Google and found a reference to problems running the 64 bit version of Bookworm. I tried the 32 bit version, and still ran into problems. After several hours of trying to use Gemini, I gave up and installed Buster using 2021-05-07-raspios-buster-armhf-full.zip. Using my notes from my Pi Zero 2 W installation, I had the Pi 4 up and running in about an hour.