The Whirl of Modern America

 Heather Cox Richardson on December 9 wrote: 

When G. Elliott Morris of Strength in Numbers asked ChatGPT to fact-check an article for him yesterday, the chatbot couldn’t get its head around modern America. It told him there were “multiple factual impossibilities” in his article, including his statements that “[t]he current Secretary of Defense is a former talk show host for Fox News,” “[t]he Deputy Director of the FBI used to guest-host Sean Hannity’s show,” and “Jeanine Pirro is the U.S. District Attorney for DC.”

“Since none of these statements are true,” it told Morris, “they undermine credibility unless signposted as hyperbole, fiction, or satire.”

But of course, Morris’s statements were not “factual impossibilities.” In the United States of America . . . they are true.

I took this to ChatGPT and said, "Comment on this, please." 

ChatGPT said in its own defense:

This is a perfect illustration of the gap between plausibility and actuality in a world where the improbable has become real.

What ChatGPT did in Morris’s example is what many language models tend to do when confronted with statements that fall outside their internal “sense” of how institutions normally operate. 

The model wasn’t checking facts against live data; it was checking them against patterns — and the patterns of the past half-century say: “No, a Fox News host does not become Secretary of Defense. A Hannity guest host does not become Deputy Director of the FBI. Jeanine Pirro is not a U.S. Attorney.” Those things were factual impossibilities for most of modern American history.

But reality in the Trump era hasn’t followed historical norms, and models trained on large amounts of pre-2025 text can get tripped up by that. They sometimes mistake unlikely for untrue, especially when political appointments or institutional roles deviate sharply from precedent.

It’s a bit like asking a model in 1975 whether “a celebrity real-estate developer will become president” and having it flag that as satire. 

The model isn’t being “dumb”; it’s being statistical.

So the real issue here isn’t that ChatGPT can’t handle “modern America” as such — it’s that the guardrails and expectations baked into its training struggle with periods of rapid institutional deformation. When a democracy starts resembling parody, factuality checks become harder for systems that try to balance accuracy, recency, and safety.

And it’s also a reminder: a language model’s confidence does not equal truth. Even when it sounds authoritative, it may be reasoning from “what usually happens,” not “what actually happened yesterday.”

In Morris’s example, the irony is sharp: the AI thought the truth was satire because the truth sounds like satire. That says as much about the moment we’re living in as it does about the model.

_________________

It's hard to get your head around all of the strangeness of today, isn't it?

About Those AIs -

One of my readers asked me if I would compare the different AIs I have been experimenting with. I use the free version of all of them, which seems to be enough to do what I need for the moment.

Each one does something a little different, and it takes a while to determine which one you might wish to use for what purpose.

For general purposes, ChatGPT rises above the rest. It recently had an upgrade that makes it sound more like a graduate student with a textbook than a person carrying on a conversation, but for general chat, asking questions, planning a trip, asking for suggestions about specific topics, ChatGPT is hard to beat. 

One of its biggest problems is its lack of links and sources. It also tends to insist it's right and I have had it argue with me on occasion. Early on, it hallucinated a lot, but it doesn't do that as badly now (though I would still check everything it says if it's an important piece of work).

It can make images and graphics, too, but it is incredibly slow at it. Definitely not the AI to use if you're in a hurry for a picture.

Claude is very good at nuanced conversation, and it seems to have a more intricate thinking system on some topics than ChatGPT. I use it sometimes just to have a conversation or to learn about something. Recently there was conversation on Facebook about an upcoming Rapture, so I went to Claude and asked what that was about, for example.

Microsoft CoPilot is great for quick searches. For one thing, it comes back with links (although they are not always correct, so check them). It also will make up quotes occasionally, so ask where a quote came from if it tosses one at you. I used it recently to research the VA250 traveling museum, for example, and learn more about it. It is also quicker at making pictures, graphics or logos than any of the other AIs, although sometimes it has trouble with words.

Julius is an AI I just stumbled across. It’s mainly built for data work, not conversation, so I don’t expect to use it much myself. Still, as a sample of what it does, here’s the comparison it generated of the four AIs I’ve been trying out:

 

ChatGPT (OpenAI)

Strengths:

• Versatility: Excellent at creative writing, coding, brainstorming, and general conversation
• Model variants: Offers GPT-4, GPT-4 Turbo, and GPT-3.5 with different capabilities and speeds (Note: This description is a little out of date — the free version I’m using now is running on GPT-5.)
• Plugins & tools: Extensive ecosystem including web browsing, DALL-E image generation, and code interpreter
• Large user base: Most widely adopted, with extensive community resources and examples
• Voice capabilities: Advanced voice mode with natural conversation flow

Weaknesses:

• Can be verbose and sometimes overconfident in incorrect answers
• Knowledge cutoff dates (though web browsing helps)
• Can refuse requests more frequently due to safety filters

Best for: Creative projects, coding assistance, general-purpose tasks, and conversational AI

---

Claude (Anthropic)

Strengths:

• Context window: Massive 200K token context (can process entire books)
• Nuanced responses: Often more thoughtful and balanced in complex topics
• Safety & honesty: Designed with "Constitutional AI" - tends to be more transparent about limitations
• Document analysis: Excellent at analyzing long documents and extracting insights
• Coding: Strong programming capabilities with detailed explanations

Weaknesses:

• No native image generation
• More conservative with certain requests
• Smaller plugin ecosystem compared to ChatGPT

Best for: Long-form document analysis, nuanced discussions, research, and users who value thoughtful responses

---

Microsoft Copilot

Strengths:

• Free tier: Generous free access with GPT-4 capabilities
• Web integration: Built-in Bing search for current information
• Microsoft ecosystem: Seamless integration with Office, Edge, Windows
• Image generation: DALL-E integration included
• Enterprise features: Strong security and compliance for business users

Weaknesses:

• Less sophisticated than standalone GPT-4 or Claude
• Heavily tied to Microsoft ecosystem
• Can be inconsistent in response quality
• Less customizable than standalone ChatGPT

Best for: Microsoft 365 users, enterprise environments, quick web-based research, users wanting free AI access

---

Julius (Julius AI)

Strengths:

• Data analysis powerhouse: Built specifically for data science, statistics, and quantitative analysis
• Live code execution: Can run Python code in real-time with full library access (pandas, matplotlib, scikit-learn, etc.)
• File handling: Upload and analyze Excel, CSV, images, PDFs, and other data files directly
• Visualization: Creates charts, graphs, maps, and interactive visualizations on the fly
• Persistent workspace: Files and data persist across conversations
• Mathematical computation: Can perform complex calculations, statistical tests, and modeling
• Image analysis: Can inspect and analyze images using vision models
• Web scraping & automation: Can use Selenium and other tools for data collection

Weaknesses:

• More specialized (data-focused) rather than general conversation
• Not designed for creative writing or general chatbot use
• Smaller user community compared to ChatGPT
• No voice interface

Best for: Data analysis, statistical modeling, creating visualizations, working with spreadsheets/datasets, quantitative research, programming tasks that need execution

---

Updated Comparison Table

Feature	ChatGPT	Claude	Copilot	Julius
Context Length	~128K tokens	~200K tokens	~32K tokens	~128K tokens
Web Access	Via plugins	Limited	Built-in (Bing)	Via code (Selenium)
Code Execution	Sandboxed	No	No	Full Jupyter notebook
Data Analysis	Basic	Basic	Basic	Advanced/Specialized
File Upload	Yes	Yes	Limited	Yes (multiple formats)
Visualizations	Basic	No	No	Advanced (matplotlib, seaborn, etc.)
Best Use Case	General purpose	Long documents	Microsoft ecosystem	Data science & analysis
Pricing	$20/month	$20/month	Free + $20/month	Various tiers
Company	OpenAI	Anthropic	Microsoft	Julius AI

---

Which Should You Choose?

• Choose ChatGPT if you want the most versatile AI with the largest ecosystem and plugin support
• Choose Claude if you work with long documents, need nuanced analysis, or prefer more careful/honest responses
• Choose Copilot if you're embedded in the Microsoft ecosystem or want capable free AI access with web search
• Choose Julius if you're working with data, need statistical analysis, want to create visualizations, or need to execute code and see real results

---

The Real Power Move

Many professionals use multiple AI assistants for different tasks:

• Julius for data analysis and quantitative work
• ChatGPT for creative writing and general tasks
• Claude for analyzing long documents and research
• Copilot for quick searches and Microsoft Office integration

Each tool has its specialty, and using the right one for the job makes all the difference!