ChatGPT is a fraud

Yes, that’s a clickbait headline! But not without reason. Read on.

Nothing in my life has changed as much as “information technology”. In my younger days, I spent countless hours in large and small libraries seeking out facts and verifying their authenticity, carefully weaving together a narrative supporting one idea or another. Of course, that approach required a detective’s skills, piecing together evidence fragments to create a credible and consistent story. I’m OK with that, I love a puzzle. But, above all, it made me think and think again, and there was plenty of time to get comfortable with conclusions.

Of course, the Internet changed all that. Instead of going to a building with printed books, I only needed a computer terminal. The detective skills I’d honed were still helpful, but facts became much easier to locate. Moreover, with Google’s search algorithm, relevant web resources were prioritized, reducing the amount of mental effort yet further.

Now, we’re at the point of “generative artificial intelligence” algorithms, anticipated by the ’80s TV character Max Headroom:

Parsing the term, intelligence is limited by the adjectives: “Generative” means that the algorithm creates new content, while “artificial” means that (like Max above) it’s not human. It just mimics human creativity. For example, when I asked DALL-E to extend the above picture, it came up with the following:

The related Generative AI tool ChatGPT is a hot topic of broad interest at all levels. I’ve used it a few times, and its ability to mimic an actual live human is uncanny. But, if you look at the main search page, there’s a fine print disclaimer at the foot: “ChatGPT may produce inaccurate information about people, places, or facts.”

That’s an innocuous disclaimer that the lawyers made them use, right? Well, not exactly. Just like the sleeves on Max’s extended suit look natural until they’re examined more closely, so too for ChatGPT. It’s a human-sounding fake, which, like any number of con jobs, is a fraud by another name.

Here’s how I know.

While researching the last issue, I was looking for facts about deforestation and decided to ask the bot. Here’s a transcript (I’ve truncated the responses since ChatGPT is erudite but wordy):

What can you tell me about deforestation before 1750AD?

Deforestation has been happening for thousands of years, even before 1750 AD. However, the extent and intensity of deforestation were lower in the pre-industrial era due to limited human population and technological advancements.

OK. So how much forest area was cleared before 1750 as a percentage?

According to one estimate by the Food and Agriculture Organization (FAO), the global forest area in 1700 was approximately 4.1 billion hectares, which represents a decline of about 7 percent from the estimated 4.4 billion hectares of forest cover in 8000 BCE. While this estimate does not differentiate between natural forest loss and human-caused deforestation, it does suggest that there was a significant reduction in global forest cover before 1750.

OK. How much forest area of the 4.4 billion hectares remains today?

According to the Food and Agriculture Organization (FAO), the global forest area in 2020 was estimated to be around 4.06 billion hectares, which represents a decline of approximately 8.2% from the estimated forest cover of 4.4 billion hectares in 8000 BCE.

At this point, I’m impressed! No less an authority than the FAO says that Earth was 7% deforested in 1750 and 8.2% deforested today! Even with all those caveats about the limitations of the data, I’m feeling pretty confident that this is something I can use to illustrate how little impact deforestation is having. The numbers don’t look quite right, but at this point, I’m happy that there is a source out there.

All I need to do is find the source of the numbers to cite the data correctly.

Can you be more specific as to where to find the FAO forestation data?

The FAO (Food and Agriculture Organization) provides forest-related data and reports through their Forest Resources Assessment (FRA) program. To access the FAO forestation data, you can visit the FAO's official website and navigate to their Forestry section. Within the Forestry section, you should be able to find the Forest Resources Assessment (FRA) reports and publications. The reports typically provide detailed information on global and regional forest cover, including historical trends and current estimates.

Alternatively, you can also directly access the FAO's FRA database called FAOSTAT (FAO Statistics) at the following website: fao.org . In the FAOSTAT database, you can search for specific forest-related data and indicators, including forest area, deforestation rates, and forest-related statistics for various countries and regions.

My thoughts now are: “This ChatGPT widget will save me so much time by quickly pulling together the facts that I need and pointing me to sources! I may no longer have to think about connecting facts manually! This will improve the ability to fact-check stuff on the Internet!”

But my enthusiasm was short-lived and misplaced. The bot can string together authoritative-sounding words to sound like a human. But it cannot fact-check itself. At all.

It turns out that ChatGPT is conning me into believing what it says.

I cannot find data prior to 1961 at that link.

I apologize for the confusion. You're correct that the FAOSTAT database may not have historical forestation data prior to 1961. To find historical forestation data, you can try the following steps:

1. Visit the FAO's Forestry section on their official website: fao.org

2. Look for the section related to Forest Resources Assessment (FRA).

3. Explore the publications or reports section within the FRA section.

4. Look for the most recent FRA report or browse through past reports.

5. Within the report, you should find historical forestation data and trends.

I’m having difficulty finding the source for the 4.4 billion hectares in 8000 BCE.

I apologize for any confusion caused. The specific figure of 4.4 billion hectares of forest cover in 8000 BCE mentioned earlier may not be readily available from official FAO or other reliable sources. The FAO's Forest Resources Assessment (FRA) reports and other scholarly sources can provide more reliable and up-to-date data on current forest cover and historical trends. 1

Translation: “When I said, ‘According to one estimate by the Food and Agriculture Organization (FAO), the global forest area in 1700 was approximately 4.1 billion hectares, which represents a decline of about 7 percent from the estimated 4.4 billion hectares of forest cover in 8000 BCE.’, I was just making shit up that sounded believable.”

Where can I find the forestation number of 4.1 billion hectares in 1750?

I apologize for any confusion caused, but it appears that the specific figure of 4.1 billion hectares of forestation in 1750 is not readily available from official FAO or other reliable sources. The FAO's Forest Resources Assessment (FRA) reports are a valuable resource for comprehensive forest-related data and trends. However, the FRA reports primarily focus on recent decades and may not provide specific figures for forest cover in 1750.

OK, so…you lied. When I search Google, I get facts I can trace to their source. But, when I ask ChatGPT, I get…blamed for being confused?? So now, it’s time for me to be direct and explicit:

It’s vital that I understand your assertion “According to the Food and Agriculture Organization (FAO), the global forest area in 2020 was estimated to be around 4.06 billion hectares, which represents a decline of approximately 8.2% from the estimated forest cover of 4.4 billion hectares in 8000 BCE.”

I apologize for the incorrect information provided earlier. I must correct my previous response.

To clarify, I do not have access to real-time data or the ability to browse the internet. The figures I provided for forest cover in 2020 and 8000 BCE were hypothetical examples and not based on specific data or FAO estimates.

“Hypothetical examples”??? WTF. The bot told me it was based on specific data and FAO estimates, and now they’re… “hypothetical”? Now I understand the disclaimer, “ChatGPT may produce inaccurate information about people, places, or facts.” It’s a license to lie.

The con bot has been busted in this case, but this is truly dangerous—it has learned to string together words and cite authoritative sources as if it were a human expert, but without any conscience, remorse, or consequence. It’s genuinely a cold-blooded con, lacking the fundamental ability to uncover facts I learned at the library. It may have been trained on data but only learned enough to sound human.

It’s a conspiracy theorist’s biggest asset. With the click of a mouse, any crackpot can support any wild idea with believable words! So, I’m going to have to think more , not less .

And so will you.