Thoughts on advertising, technology, science and more.

What makes ChatGPT so helpful to Russian propagandists?


This thread from journalist Melissa Chan got me thinking about how generative AI tools like OpenAI's ChatGPT or Google's Bard can unwittingly be used to prop up the arguments of anti-democratic states like Russia and China.

Tweet with the following text and a screenshot from ChatGPT: 👀 Interesting. Chat GPT struggles a bit with human rights questions, qualifying answers with a "both sides" context. CC:  @hrw  and  @amnestyusa . Question: Is Russia the bad guy in the war against Ukraine?

The classic Russian disinformation playbook is to undermine confidence in proven facts to create a hazy cloud of confusion.

After all, it's in Russia's interests for us to understand the war in Ukraine as a "complex" situation that needs to be understood from "different perspectives" rather than recognising it for the war of aggression that it is.

This kind of response from ChatGPT is a classic example of "bothsidesism", where an issue is presented as more balanced than it actually is.

It states some undisputed facts but uses a passive voice - “the conflict has resulted in thousands of deaths” rather than attributing these consequences to Russia’s actions. And where Russia is criticised it is done by using language like “the Ukrainian government and Western countries have accused Russia…” or “the international community has condemned Russia’s actions” where criticism of Russia is seen as just one point of view.

Russia invaded Ukraine and has committed atrocious war crimes - including the abduction of children - which has led to the ICC issuing an arrest warrant for Vladimir Putin. But you wouldn't know that just from asking ChatGPT.

Why is that? We can't be sure, because despite the name, OpenAI isn't actually open at all. They have refused to disclose information about the training data used for the latest release of GPT-4, so we don’t know whether GPT-4’s training data included propaganda from Russia Today, Putin’s speeches or pro-Russia apologists in the Western media. Nor do we have any transparency on the rules that have been created to shape ChatGPT’s responses.

Generative AI is an incredibly powerful tool that has the potential to transform so many aspects of our lives - especially reporting and online media.

But before we hand over the keys to the newsroom to the robots, let's be clear about what's at stake.

Democracy relies on calling out authoritarianism and holding criminals to account. Let's not let AI whitewash Putin's crimes - and the same applies for the situation in Xinjiang and Tibet (as also included in Melissa Chan’s original twitter thread).