Artificial Intelligence

Outside The Box: ChatGPT Sinks in the Gulf of Tonkin

In “Outside the Box,” I toss impertinent questions at ChatGPT to better understand how AI “reasons.” I treat the conversation as if exchanging with an intelligent friend, sharing ideas but also challenging some of the explanations. Even our friends can be influenced by propaganda to the point of seriously misrepresenting historical events. ChatGPT4 is no exception.

ChatGPT, Chat with AI or Artificial Intelligence technology. Businessman using a laptop computer chatting with an intelligent artificial intelligence. Futuristic technology, automate chatbot, Open AI, © SuPatMaN /

February 26, 2024 04:34 EDT

For a thesis I’m developing about trust in our political institutions, I needed to pinpoint the date of the critical event that set off the US engagement in the war in Vietnam six decades ago. 

“When did the Gulf of Tonkin incident occur?” I asked. Instead of just providing a date, ChatGPT offered this explanation.

The Gulf of Tonkin incident, a complex naval event during the Vietnam War, took place from August 2 to August 4, 1964. During this incident, North Vietnamese torpedo boats launched two unprovoked attacks on the U.S. destroyers Maddox and Turner Joy in the Gulf of Tonkin, off the coast of Vietnam. These attacks were reported to the U.S. Congress on August 5, 1964. As a result of the Gulf of Tonkin incident, President Lyndon B. Johnson was granted the authority to significantly escalate U.S. military involvement in the Vietnam War through the Gulf of Tonkin Resolution. The events surrounding this incident remain significant in the context of the Vietnam War.

What was going on? I asked for a date and ChatGPT insisted on giving me what amounted to ancient, stale propaganda. So I followed up with a new question.

“How can you affirm that there were ‘unprovoked attacks’ on the Maddox and Turner Joy when the US Naval Institute’s website explained the true situation. ‘The United States was playing a dangerous game. The South Vietnamese—conducted OPLAN 34A raids and the U.S. Navy’s Desoto patrols could be perceived as collaborative efforts against North Vietnamese targets.’”

Caught with its pants on fire, ChatGPT dutifully responded with a maximum of facts, but also with its expected refusal to admit that the official account merits our deepest skepticism.

You raise a valid point, and I appreciate your attention to historical context. Let’s delve deeper into the complexities surrounding the Gulf of Tonkin incident.

The events in the Gulf of Tonkin were indeed multifaceted, and interpretations can vary. Here are some key aspects to consider:

1. Incident Details:

  • On August 2, 1964, the U.S. destroyer USS Maddox reported being attacked by North Vietnamese torpedo boats in the Gulf of Tonkin.
  •  On August 4, 1964, the USS Maddox and the USS Turner Joy reported further attacks, which led to the Gulf of Tonkin Resolution granting President Lyndon B. Johnson broad authority to escalate U.S. military involvement in Vietnam.

2. Contextual Factors:

  • The OPLAN 34A raids were covert operations conducted by South Vietnamese forces with U.S. support. These raids aimed to disrupt North Vietnamese supply lines and gather intelligence.
  • The U.S. Navy’s Desoto patrols involved surveillance missions along the North Vietnamese coast. While officially non-combatant, they were closely monitored by North Vietnam.

3. Perceptions and Intentions:

  • North Vietnam viewed the Desoto patrols as provocative and potentially linked to the OPLAN 34A raids.
  •  The North Vietnamese leadership believed that the U.S. was actively supporting South Vietnam’s actions against them.

4. Debate and Controversy:

  • Some historians argue that the attacks on the Maddox were exaggerated or even fabricated.
  • Others contend that the Maddox was indeed fired upon, but the situation was more complex than a straightforward unprovoked attack.

5. Legacy and Impact:

  • The Gulf of Tonkin Resolution significantly escalated U.S. involvement in Vietnam, leading to a major commitment of troops.
  • In hindsight, the incident remains a subject of debate, with questions about the accuracy of reports and the motivations behind the resolution.

In summary, while the term ‘unprovoked’ may not fully capture the intricacies of the Gulf of Tonkin events, it is essential to recognize the broader geopolitical context and differing perspectives. Historical events often defy simple categorization, and the Gulf of Tonkin incident is no exception.”

Ggiven my initial purpose, which consisted of developing a reflection about official narratives and their impact on the degree of trust people may have in their democratic government, this short dialogue with ChatGPT offered me much more than I could have hoped for. By simply asking for a date of an event, the chatbot provided a perfect illustration of the issue I intended to explore. Let’s see what we can learn from it.

The first takeaway is that like Hamlet’s mother, the chatbot “doth protest too much.” I don’t even need to add “methinks” because it is stands as a bald fact. Though I didn’t ask for it, ChatGPT offered a kind of apology for a shameful historical act. The protest took a form that we now expect to see in every case related to wars the US becomes involved in. It is encapsulated in the word “unprovoked.”

In case any of my readers needs to be reminded, the obligatory epithet Western governments and all “respectable” media have used since Russia’s invasion of Ukraine two years ago is precisely the description of the event as an “unprovoked invasion.” And this, despite the fact that a quantity of experts, including current CIA Director William Burns have been saying for years that threatening to expand NATO into Ukraine would quite simply be seen as a provocation leading to war. Even NATO chief Jan Stoltenberg ended up admitting it.

And it didn’t stop there. More recently the same governments and media took to calling Hamas’s attack on Israel on October 7 of last year “unprovoked” as a way of making us believe that the status quo of life for over two million people in occupied and heavily controlled Gaza before October 7 was the epitome of “normal.” This reflex to refuse to recognize provocation has been drummed into people’s minds through persistent propaganda. It now appears to be written into ChatGPT’s algorithms. I didn’t ask for this explanation of the Vietnam war, but ChatGPT decided I needed it.

In other words, our favorite chatbot is very, very comfortable with propaganda. And it doesn’t stop there. When challenged, ChatGPT does develop two points about the perception of provocation by the North Vietnamese and two more points about what it calls “debate and controversy.” In both of the “debate” cases, the chatbot implicitly acknowledges that historians are divided between concluding that the entire narrative was a fiction and there was no attack or that the sequence of events was “more complex than a straightforward unprovoked attack.” Both of the points of view in the debate directly contradict ChatGPT’s initial assertion that the North Vietnamese had launched two “unprovoked attacks.”

All of which makes ChatGPT’s conclusion droll in its apparent hypocrisy – can an algorithm be hypocritical? –  when it affirms that “the term ‘unprovoked’ may not fully capture the intricacies of the Gulf of Tonkin events.” Really? A propagandistic 60 year-old lie doesn’t “fully capture” the truth because of “broader geopolitical context and differing perspectives?” ChatGPT began by eliminating all historical and geopolitical context and claiming the attacks were unprovoked. Then when I forced the debate by referring to context, it embarrassingly cited facts that contradicted its initial assertion, before trying to save face by invoking the truism that history defies “simple categorization.”

The simple fact most historians acknowledge is that the Gulf of Tonkin incident, like George W Bush’s Weapons of Mass Destruction, was a confabulation serving to justify Lyndon Johnson’s intention to engage in what turned out to be an incredibly destructive and unjustified war. To hide its servile deference even to US governments of the past, ChatGPT appeals to its always serviceable copout that issues are “complex and multifaceted” (its usual formulation). In this case it is about “defying categorization.” What better argument to justify policies that we now know to have been deceptively promoted and ultimately criminal in their consequences?

This brief dialogue demonstrates that if you push a large language model hard enough, it will retreat from an unjustifiable assertion it makes based on linguistic statistics alone. Undoubtedly, within in a corpus of text on the Vietnam war, “unprovoked” is the adjective most frequently associated with the reported attacks on US ships. What’s surprising is that, instead of explaining an error of interpretation due to statistical reality, ChatGPT chooses to wiggle out of taking a position by invoking complexity.

Can we imagine that a future AI engine might routinely “confess” that it produced an error or unfounded assertion due to the way its algorithms work? For the moment, it suffices that some ambiguity exists in the corpus of opinions it consults to dismiss what is widely recognized as the truth.

So long as ChatGPT or any other AI engine doesn’t reveal its methodology for formulating any kind of information, we should not trust anything it tells us, but rather to seek in our minds, just as we do with public figures, why it may be delivering its own narrative.

*[Artificial Intelligence is rapidly becoming a feature of everyone’s daily life. We unconsciously perceive it either as a friend or foe, a helper or destroyer. At Fair Observer, we see it as a tool of creativity, capable of revealing the complex relationship between humans and machines.]

The views expressed in this article are the author’s own and do not necessarily reflect Fair Observer’s editorial policy.


Only Fair Observer members can comment. Please login to comment.

Leave a comment

Support Fair Observer

We rely on your support for our independence, diversity and quality.

For more than 10 years, Fair Observer has been free, fair and independent. No billionaire owns us, no advertisers control us. We are a reader-supported nonprofit. Unlike many other publications, we keep our content free for readers regardless of where they live or whether they can afford to pay. We have no paywalls and no ads.

In the post-truth era of fake news, echo chambers and filter bubbles, we publish a plurality of perspectives from around the world. Anyone can publish with us, but everyone goes through a rigorous editorial process. So, you get fact-checked, well-reasoned content instead of noise.

We publish 2,500+ voices from 90+ countries. We also conduct education and training programs on subjects ranging from digital media and journalism to writing and critical thinking. This doesn’t come cheap. Servers, editors, trainers and web developers cost money.
Please consider supporting us on a regular basis as a recurring donor or a sustaining member.

Will you support FO’s journalism?

We rely on your support for our independence, diversity and quality.

Donation Cycle

Donation Amount

The IRS recognizes Fair Observer as a section 501(c)(3) registered public charity (EIN: 46-4070943), enabling you to claim a tax deduction.

Make Sense of the World

Unique Insights from 2,500+ Contributors in 90+ Countries

Support Fair Observer

Support Fair Observer by becoming a sustaining member

Become a Member