After the adventures of last week, I started this writing adventure wanting to try to figure out what people were doing with LangChain and social media. People are both generating content for social media using LLMs and oddly enough repurposing content as well. We have to zoom out for just a second and consider the broader ecosystem of content. In the before-times, people who wanted to astroturf or content farm had some work to do within the content creation space. Now ChatGPT has opened the door and let the power of synthetic content creation loose. You can create personas and just have them generate endless streams of content. People can even download and run models trained for this purpose. It’s something I’m legitimately worried about for this next election cycle. Sometimes I wonder how much content within the modern social media spaces is created artificially. Measuring that is actually pretty difficult. It’s not like organically created content gets a special badge or recognition.
For those of you who were interested in finding out insights on any topic with a plugin that works with the OpenAI ChatGPT system then you could take a moment and install “The Yabble ChatGPT Plugin” [1]. Fair warning on this one I had to reduce my 3 plugins down to just Yabble and be pretty explicit in the prompts within ChatGPT to make it do some work. Sadly, I could not just login to Yabble and had to book a demo with them to get access. Stay tuned on that one to get more information on how that system works. I had started by searching out plugins to have ChatGPT analyze social media. This has become easier now with the announcements that OpenAI can openly use Bing search [2].
Outside of searching using any OpenAI tooling like ChatGPT, Google was pretty clear on the reality that what I was really looking for happened to actually be marketing tools. Yeah, I went down the SEO Assistant rabbit hole and it was shocking. So much content exists in this space that is like watching a very full ant farm for the most part. Figuring out where to jump in without getting scammed is probably a questionable decision framework. Whole websites and ecosystems could be synthetically generated pretty quickly. It’s not exactly one click turn key deployments, but it is getting close to that level of content farming.
I was willing to make the assumption that people who were going to the trouble of making actual plugins for ChatGPT within the OpenAI platform are probably going to be more interesting and maybe are building actual tooling. For those of you who are using ChatGPT with OpenAI and have the plus subscription you just have to open a new chat, expand the plugin area, and scroll down to the plugin store to search for new ones…
I also did some searches for marketing tools. I’m still struck with the possibility that a lot of content is being created and marketed to people. It’s not the potential flooding of content that becomes so overwhelming that nobody is able to navigate the internet anymore. We are getting very close to the point where it would be entirely possible for the flooding of new content to occur in ways that simply overwhelm everybody and everything with new content. This would be like the explosion of ML/AI papers over the last 5 years, but maybe 10x or 100x even that digital content boom [3].
Footnotes:
[1] https://www.yabble.com/chatgpt-plugin
[2] https://www.reuters.com/technology/openai-says-chatgpt-can-now-browse-internet-2023-09-27/
[3] https://towardsdatascience.com/neurips-conference-historical-data-analysis-e45f7641d232
What’s next for The Lindahl Letter?
Week 144: Knowledge graphs vs. vector databases
Week 145: Delphi method & Door-to-door canvassing
Week 146: Election simulations & Expert opinions
Week 147: Bayesian Models
Week 148: Running Auto-GPT on election models
If you enjoyed this content, then please take a moment and share it with a friend. If you are new to The Lindahl Letter, then please consider subscribing. New editions arrive every Friday. Thank you and enjoy the week ahead.
Synthetic social media analysis