Kenyans drag their president for allegedly sending 100% AI-generated congratulatory letter to Trump
A section of Kenyan X users are dragging their president, William Ruto, for allegedly generating his congratulatory message to US President-Elect, Donald Trump in its entirety with Chat GPT4.
An X user, @MrBoenKE took to the platform to publish what he purported to be the result of a tech-enabled assessment of the congratulatory letter which shows that the letter has zero human input.
The claim went viral with a lot of Kenyans bashing Ruto and his communication team for tardiness.
However, an independent check of the same letter by The Kenya Times suggests that this claim may not be true.
The publication used various AI detectors to fact-check the claims and published its findings today, Thursday, November 7, 2024.
The publication published the result of a test conducted with Isgen detector, showing that the text was entirely generated by a human with zero probability that AI was used.
Other tools used included Copyleaks AI detector and GPTZero, which also showed a zero per cent probability of the text being AI-generated.
Quetext AI detector gives a tiny percentage of 0.012 to AI, and affirms that: “Our detector is highly confident that the text is written entirely by a human.”
However, Kenya Times’ fact-checking exercise became complicated when they ran the text through more tools.
The publication said Free AI Detector & ChatGPT Detector results suggested that there are chances that the text was generated using AI.
“There is relatively high risk that your text was written by AI,” reads the results.
According to the tool, 87 per cent of the words in the text are commonly used by AI generators when creating content.
In the breakdown, 9 per cent of the words are occasionally used by AI writers while 3 per cent can’t be probably used by AI writers.
Only 1 per cent of the words Ruto used are unique and unpredictable to be incorporated by an AI.
According to the tool, 70 per cent is human content while 30 per cent could have been generated using AI.
Quillbot and Sribbr showed that 48 per cent of the text is likely to be AI-generated with only 58 per cent being human content.
Grammarly’s AI detector says 25 per cent of the text appeared to be AI-generated.
