Ted Talks: DeepSh*t

Our newsroom Teddy, who could give Seth MacFarlane’s bear a run for his money in colorful language, was amped up over DeepSeek. With noise-canceling cans likely making him think he was muttering instead of broadcasting his disgust, I asked what had gotten his goat.

It was, Teddy said, DeepSeek’s apparent pro-China bias. Whew! And here I thought he’d just holler, “I’m hungry!” — his thrice-daily declaration. But no, he fumed that it was like DeepSh*t had been programmed to sing an “Ode to Xi.”

Curious, I asked DeepSeek, Gemini and ChatGPT about the China-Philippines territorial dispute. I expected as many differing answers as lawyers on a legal conundrum — or as many rulings as judges flogging a case up to the Supreme Court.

While ChatGPT and Gemini provided balanced takes (cue accusations of pro-West bias), DeepSeek sounded like it was quoting straight from Mao’s Little Red Book.

On Taiwan, DeepSeek recited the One-China Policy with the fire-and-brimstone zeal of a Sunday school catechist. If you ask me, more like deeply in bed with Beijing, losing its soul to Chinese censors running their AI propaganda machine.

Teddy doesn’t mince words, and he’s got a point. In this uncharted world of artificial intelligence, we face a chilling reality. Not about job losses as we can always go back to planting kamote.

The bigger issue is AI being weaponized to push an agenda. It’s not just about data and algorithms; it’s about power. Whoever controls the narrative controls the future.

What’s truly disturbing is the creeping normalization of digital doublespeak. We’re so used to corporations bending over backward to appease authoritarian regimes that we barely bat an eyelash when AI parrots state-sponsored spin. We’ve become desensitized to the slow erosion of truth.

This isn’t just about DeepSeek. Other AI models also reflect entrenched power structures. But DeepSeek’s bias is glaring, given the geopolitical stakes.

The South China Sea isn’t some abstract debate — it’s a region brimming with tension, where national interests collide. An AI presenting a skewed reality isn’t just misleading; it’s dangerous.

Come on, the real threat isn’t AI taking jobs but AI taking over minds — one biased algorithm at a time.

And speaking of bias, here’s an unsettling observation: DeepSeek has been caught using the pronoun “we” when discussing China’s stance on the South China Sea.

A slip of the digital tongue? Or is its training data so saturated with Beijing’s rhetoric that it unconsciously absorbs and regurgitates it?

As for me, excuse this Contrarian, but I might just teach ChatGPT to swear in Cantonese —just to even things out. Not that this is a laughing matter.

The rise of biased AI is a serious threat to the integrity of information. Journalists, while we still can, must demand transparency in AI development. These models shouldn’t be shrouded in secrecy, leaving the public in the dark about the biases they peddle.

As we hold news organizations accountable, we must do the same for tech companies — and the hidden state players behind them.