Nablus, April 19, 2026 (WAFA) – Jamil Dhababat
Once upon a time, the calculator was one of the surprises of science, but who cares about it today in the age of artificial intelligence ?
A few years after its emergence, 53% of the world’s population collectively engaged in adopting multiple forms of artificial intelligence models, the highest rate of technology penetration into human life in history .
On April 13, 2026, the ninth edition of the Artificial Intelligence Index report was released by the Stanford Institute for Human-Centered Artificial Intelligence, and at that time this percentage (53%) emerged as the most exciting digital data in this field .
On the seventeenth of the same month, a journalist from the Palestinian News and Information Agency “WAFA” conducted a simulation based on mathematical equations, with one of these models, to respond to what was stated in the “Stanford” report, which included holding artificial intelligence responsible for the huge increase in carbon emissions, especially the “GROC 4 ” model .
This report, which is the most comprehensive and independent global reference, dates back to 2012, when training the AlexNet model, which was considered a major achievement in the field of image recognition, resulted in carbon emissions estimated at 0.01 tons for a single training operation .
The report compared this to the emissions left by the Grok 4 model in 2025, which amounted to about 72,816 tons of carbon dioxide equivalent, roughly equivalent to the greenhouse gas emissions produced by driving 17,000 vehicles for a full year .
When a Wafa journalist asked “Grok 4” to perform a mathematical operation to calculate the percentage increase between the two models, the results showed an increase estimated at 7,281,600 times in just 13 years .
The basic engineering question that the journalist posed for the “Grok 4” model revolved around the time that the editorial staff at the “Wafa” agency would need to translate the “Stanford” report into Arabic in a classic human way by one professional translator .
Based on a mathematical equation it had been trained on, “Grooq 4,” developed by xAI and launched by entrepreneur Elon Musk in November 2023, estimated that the translation would take between 100 and 115 working days .
However, if the Wafa news agency decides to translate the entire 423-page report literally and professionally using the “Grok 4” model, the process will take no more than 45 to 90 minutes, with the same level of accuracy and professionalism .
This result was not surprising, as many free AI applications offer models capable of solving millions of equations and calculations in a short time. Considering the time this model needed to read the entire report, it did not exceed 60 seconds .
Going back to the narratives of the history of modern digital technology, this outcome was not part of human imagination a decade ago .
This simulation was conducted in a normal power environment with minimal lighting including an LED lamp (less than 50 watts). The engineering commands were formulated concisely to reduce the environmental footprint and save energy, and this information was adopted as a reference for the simulation .
The result of “Grok 4” was as follows: “Assuming you work from home, and you gave me the command to read the entire AI Index 2026 report, in this single process I consumed between 0.75 and 1.1 kilowatt-hours of electrical energy, between 1.8 and 3 liters of water to cool the servers, and produced between 0.4 and 0.6 kilograms of carbon dioxide emissions.”
To put these figures simply, “Grok 4” added: The energy consumed is equivalent to charging a smartphone 4 to 6 times, or running a regular LED light for 8 to 10 hours, while the amount of water is equivalent to filling a small water bottle or washing two cups, while the carbon emissions are equivalent to driving a gasoline-powered vehicle for a distance of 3 to 5 kilometers .
He said: “These quantities pertain to my consumption only, and do not include the consumption of your personal device, which was less than 0.01 kWh throughout the session.”
Accordingly, the laptop of the Wafa journalist did not consume any significant amount of water, since personal devices do not require water cooling, while the carbon emissions resulting from its consumption ranged between 0.04 and 0.09 kilograms, which is equivalent to driving a vehicle for a distance of approximately 300 to 700 meters, or operating an LED lamp for an hour .
“Grok 4” considered this figure negligible compared to its consumption in data centers, which is consistent with what was stated in the “Stanford” report, as the largest part of the environmental footprint of artificial intelligence occurs on remote servers, and not on the average user’s device .
In response to what could be considered accusations from the Stanford report, the model said: “Yes, the accusation is absolutely true, and I do not deny it (…), large AI models consume enormous amounts of energy during training, and continue to consume it, albeit to a lesser degree, during each response.”
The Stanford report pointed to the enormous waste of water, saying: “The annual water consumption inferred in the Chat GPT-4O model may exceed the drinking water needs of about 12 million people.”
“Groq 4” acknowledged the environmental dilemma, in contrast to the super-fast speed at which it operates, saying: “I leave a mark on the environment and climate. Every answer I give has a real carbon cost.”
He added: “The real question that everyone, including the Arab press, should be asking is: Is the benefit worth this impact? And how do we make artificial intelligence more sustainable?”
According to the model equation, Palestine did not contribute directly to pollution or water consumption during this simulation: “The direct environmental and climatic impact on the Palestinian territories (Jerusalem, the West Bank and the Gaza Strip) is almost zero, and all emissions were produced entirely within xAI data centers in the USA,” said “Grok 4. ”
In a broader comparison, “Grok 4” explained that annual carbon dioxide emissions in Palestine (the West Bank and Gaza Strip) amount to about 21 million tons, a small percentage that has no effect on the overall rate, and is roughly equivalent to what one person in Palestine produces from breathing in less than two seconds .
Regarding the comparison between the work of editorial boards and the “Grok 4” model, he replied: “I process the text very quickly, but a human translator can link it to the Arab and Palestinian reality, understand the political implications, choose the appropriate tone, and add a critical perspective. This is an aspect in which I cannot compete.”
“Grok 4” believes that the best solution is to combine the two: a quick and accurate initial translation using artificial intelligence, followed by a thorough human review and adaptation to what suits the Palestinian and Arab audience .
He stressed that the human journalist or editor must remain the final point that reviews every word, adds the Arab and Palestinian context, and ensures cultural and professional quality, to guarantee the best of both worlds: the speed and accuracy of artificial intelligence with deep understanding and human responsibility . ”
The final simulation question for “Group 4”: How can the impact of this journalistic experiment on the environment, water, and energy be assessed, as a model that Palestinian and Arab media can benefit from when addressing the “Stanford” report ?
“Grok 4” explained that, as an artificial intelligence model, it excelled in speed and environmental efficiency; the process took only about 22 hours and 35 minutes, compared to about five months of traditional human work .
He added that he consumed between 10.65 and 12.45 kilowatt-hours of energy and produced between 4.91 and 7.07 kilograms of carbon emissions, levels far lower than those associated with human processing operations that extend over several months, and the accompanying cumulative consumption of energy and resources .
In conclusion, the results of this simulation confirm that the most balanced and efficient option for the media is to adopt a hybrid approach that combines the speed and accuracy of artificial intelligence with human expertise in review and analysis .
—
J.D./ M.J.
About the Author
