Fact-Checking
Although ChatGPT and other AI systems have received massive media attention since late 2022, I only began dipping my toe in those waters about a year ago.
At that time, I released a series of chatbots for many of the individual authors on our website, with each of these produced by focusing it individually upon the total content we’d published from that corresponding writer.
These notably included the Ron Unz Chatbot, drawing upon roughly a million and a half words of my own articles, and some of our readers have subsequently used it to simulate my own reactions on various issues. For example, earlier this month a commenter used this Unz chatbot to write a 10,000 word article on some aspects of McCarthyism, which he published as a very long comment.
- American Pravda: Laminated Mouse-Brains and the Magic of AI
Ron Unz • The Unz Review • June 10, 2024 • 15,000 Words - Questioning Our Writers, Whether Living or Dead
Ron Unz • The Unz Review • June 15, 2024 • 6,600 Words
Several months ago, I added another AI feature, having the ChatGPT system automatically produce short summaries and outlines for every article we published that was longer than 1,000 words. This allowed readers to easily get a rough sense of those pieces that they lacked the time or interest to actually read:
- Providing AI Summaries of Website Articles
Ron Unz • The Unz Review • February 14, 2025 • 1,400 Words
However, with hundreds of billions of dollars of ongoing capital spending on AI, these software systems have continued to rapidly improve, and I recently learned that OpenAI had released a new and especially powerful version of ChatGPT called Deep Research.
Whereas ChatGPT and most other chatbots are designed to respond within seconds, the Deep Research AI may spend up to 30 minutes working on a given assigned topic, but it uses that time to produce remarkably advanced results. For example, on a standard benchmark test, the GPT-4o system scored only 3.3%, DeepSeek’s R1 model did much better at 9.4%, but Deep Research rated a vastly superior 26.6%.
Once I began testing the Deep Research AI, these numbers seemed quite plausible to me. I discovered that the system can very effectively be used to fact-check long, complex articles of the sort that I often write. After a couple of such tests, I was so impressed that I have now had dozens of my American Pravda articles fact-checked by Deep Research.
I was hardly surprised that such a powerful new AI system was also very resource-intensive, so basic ChatGPT users have been limited to twenty-five Deep Research runs per month, with only the first ten being at full power and quality. Moreover, since the system is so new, these runs sometimes fail, with those failures still being counted towards that monthly limit.
But despite those minor inconveniences, I was absolutely astonished by the analytical quality of what Deep Research produced, results that fully validated the dramatic claims made in media accounts.
It’s been widely recognized that all of these recent AI systems have easily blown past the decades-old “Turing Test” of machine intelligence, but the output of the Deep Research AI was entirely on a different level. Many or most of its full-power analysis runs seemed as if they had been written by an exceptionally intelligent individual who had read nearly everything published on the entire Internet and also had almost total recall.
Although I’m still not entirely convinced that the hundreds of billions of dollars currently being invested in AI will ever produce an adequate financial return, the AI systems created are certainly one of the most amazing things I’ve ever encountered, being closer to magic than software technology, and doing things I never would have believed possible in a million years. If some company had invented a practical teleportation device, I probably would have regarded such a product as much less remarkable.
Regardless of any future advances, I have already found the current fact-checking capabilities of the Deep Research AI extremely useful, especially with regard to the very controversial and non-mainstream content of so many of my own articles.
As a trivial example of the power of the system, it easily caught several small factual mistakes that I’d made in some of my articles, mistakes that for many years had escaped my own notice and that of all my readers. Reacting in exemplary fashion, Deep Research flagged these as minor, careless errors that should be corrected while also dismissing the notion that they seriously detracted from the broader accuracy of those articles.
At the top of each fact-checking run, I have included a selection of the AI remarks and my own respon
Article from LewRockwell
LewRockwell.com is a libertarian website that publishes articles, essays, and blog posts advocating for minimal government, free markets, and individual liberty. The site was founded by Lew Rockwell, an American libertarian political commentator, activist, and former congressional staffer. The website often features content that is critical of mainstream politics, state intervention, and foreign policy, among other topics. It is a platform frequently used to disseminate Austrian economics, a school of economic thought that is popular among some libertarians.
