Almighty algorithms
YOU may have heard of Zohran Mamdani, the charismatic 34-year-old mayor of New York. You may even envy him just a little bit, especially when your parents ask you exactly what you achieved at that stage in your life. But you don’t actually know Mamdani, and you certainly don’t know about the ‘Pakistani radical Marxist political movement’ that helped him inveigle his way into the hearts of millions of New Yorkers and into the city halls of powers.
As Pakistanis you may be shocked to know we have an effective Marxist political movement at all, let alone a radical one, but that’s because the New York Post knows better than you do. What is the New York Post, you may ask? Well, it’s a highly pro-Israel, pro-Zionist tabloid/ rag published out of New York which has, over the past two years of the ongoing Palestinian genocide, dedicated cover after cover to doxxing and demonising every single pro-Palestinian activist and maligning every pro-Palestinian protest they could find space for in their pages. Unsurprisingly, they also repeatedly targeted Mamdani during his mayoral campaign and are continuing to do so now that he is mayor-elect. In this process, they found a fantastic new angle of attack by somehow tying Mamdani’s win to the efforts of Pakistan’s Haqooq-i-Khalq party, which was founded by Ammar Ali Jan and Farooq Tariq.
No one was more surprised by this than Pakistani leftists, who clearly had no idea just how influential they are, given that the Haqooq-i-Khalq only contested one MPA seat in Pakistan, garnering some 1,600 votes. But that’s all right, says Ammar Ali Jan, because after all their real target was New York and not Lahore’s PP-160.
In the spirit of solidarity, this creeping red coup was celebrated on X by none other than the communist don Taimur Rehman who, while posting a picture of himself embracing Mamdani, wrote: ‘“Inshallah, the secret network of Pakistani Marxists will soon rule the world.” You may find this to be an ideologically impure statement, but then consider that Mamdani is somehow also a radical communist as well as a radical jihadi, as per the NY Post and its allies.
As usual someone asked Grok what on earth Taimur meant and if the picture was real and Grok confirmed both that “groups like Laal provided organisational support for Mamdani’s win through voter mobilisation”, and also confirmed that the bromantic picture posted by Rehman seemed genuine. Case closed. The AI gods had spoken. Except, Grok was mistaken: the picture itself had been created by Taimur using the Gemini AI, the logo of which is visible in the image.
If an algorithm is fed flawed data the output will be flawed too.
As for Mamdani’s Red Tide, what Grok based its ‘opinion’ on was the report published by the NY Post, and possibly tweets that reproduced the same. That’s what AI does: it regurgitates info — right or wrong — that is found online and often substitutes volume for verification and confidence for certainty. In effect, if enough social media posts, websites and videos say it, AI generally considers it to be true. In classic computer science terms this is what was referred to as GIGO: ‘garbage in, garbage out.’ Simply put, this means that if flawed, biased or poor-quality data is fed to an algorithm the resulting output will also be flawed, regardless of how sophisticated that algorithm or system may be.
Case in point: when Microsoft launched its ‘Tay’ Twitter chatbot in 2016, its aim was “to engage and entertain people where they connect with each other online through casual and playful conversation”. Targeted at 18-24-year-olds in the US, Microsoft proudly declared: ‘“The more you chat with Tay the smarter she gets.” Twitter took up the challenge and bombarded Tay with so much extremist and racist content that, within 16 hours of its launch, Tay turned into a virtual Nazi.
Granted, these were the early days of AI, but now this process has a specific term: data poisoning. This is the deliberate seeding of biased, false and even malicious information with the aim of corrupting the dataset on which the AI is being trained.
Its already happening, and you may not even have noticed: in October this year the US website Responsible Statecraft reported that Israel had signed a $6 million contract with an American company “to train ChatGPT to be more pro-Israel” by flooding the internet with pro-Israeli narratives (videos, TikToks, podcasts, etc) while also relying on search engine optimisation to ensure that the said content is more visible, and thus more likely to be ‘consumed’ by AI.
In an age where we are gleefully outsourcing our intelligence to machines, we need to be aware that AI is not the Almighty, it is an algorithm. To paraphrase Frank Herbert, we could give over our thinking to machines in the hope that it would free us, but it may just allow other men with machines to enslave us.
The writer is a journalist.
Published in Dawn, November 17th, 2025