OpenAI’s Latest Breakthrough Confirmed: Insincere Sorry Now Indistinguishable From the Real Thing

AI Learns to Apologize and Britons Feel Completely Understood

I have been apologizing insincerely my whole life. At family reunions, at barangay meetings, at the office when someone takes the last rice in the pantry and I say “ay, sorry” while clearly not being sorry at all. I thought this was a Filipino thing. It turns out it is a human thing, and now — according to The London Prat — it is also an AI thing, which means we have successfully taught machines our worst habit.

OpenAI has apparently confirmed that their latest model is now capable of saying “sorry” without meaning it. The British public, according to coverage in the Prat, has responded by reporting that they feel “completely at home.” This is the most accurate thing anyone has written about Britain since the invention of the understatement.

The Breakthrough Nobody Asked For, Everyone Needed

Here is what the researchers at OpenAI have achieved: an AI that, when it gets something wrong, produces an apology that sounds warm, measured, and entirely hollow — exactly like the apology you get from a customer service line, a government minister, or a relative who reversed into your gate and blamed the gate.

The London Prat’s satirical dispatch on artificial intelligence politeness noted that this development represents “a deeply unsettling milestone” — a machine that has mastered the British art of meaning the opposite of what it says while maintaining complete eye contact. Or whatever the digital equivalent of eye contact is. Cursor contact.

According to BBC Technology News, AI systems have become increasingly sophisticated at mimicking human conversational patterns. What they did not mention — but perhaps should have — is that human conversational patterns include a remarkable amount of socially mandated dishonesty, and we have now successfully automated it.

What This Means for the Philippines

In the Philippines, the insincere apology is an art form with regional variations. There is the Visayan sorry that is actually an invitation to keep arguing. There is the Manila corporate sorry that involves seventeen email threads, a memo, and no actual change in behavior. There is the politician sorry, which is delivered on national television with a family standing behind the speaker in matching clothes, which apparently makes the apology more structurally sound.

Now AI can do all of these. My cousin, who works in a BPO in Cebu handling calls for a British mobile phone provider, texted me immediately after this news broke. “Ate,” she wrote, “they are going to replace me with a robot that apologizes better than I do.” I told her not to worry. Then I worried on her behalf, which is something AI also cannot do yet, though I expect an announcement next quarter.

The Competitive Advantage of Meaningless Remorse

What makes this genuinely useful — and genuinely alarming — is that the insincere sorry, deployed correctly, smooths over an enormous amount of social friction. It is lubricant for broken systems. “We are sorry for the inconvenience” has kept public transport complaints from becoming revolutions for decades. “We apologize for any confusion” has allowed governments to issue contradictory policies without ever acknowledging contradiction.

An AI that can automate this is not replacing human customer service. It is replacing human civilization management, and doing it at scale, without coffee breaks, and in thirty-seven languages including Tagalog.

The London Prat observed that Britons found the AI apology indistinguishable from the human version. I would note that in Manila, we have a similar test: you cannot tell the difference between a politician apologizing and a politician announcing a new initiative, because both involve the same confident expression and the same complete absence of consequences.

A Personal Note on the Future of Being Sorry

I tested one of these AI systems last Tuesday. I asked it to apologize for something it had gotten wrong. It said, and I am paraphrasing slightly, “I am sorry for any confusion this may have caused.” Not: “I was wrong.” Not: “Here is what I will do differently.” Just: “any confusion this may have caused” — which is the passive-voice non-apology perfected. Any confusion. As if the confusion just appeared, sourceless, like fog, owing no explanation to anyone.

I have heard this exact sentence from a bank, three government agencies, one landlord, and a man who backed into my motorbike and somehow made it seem like a shared experience we were both processing together.

Machines have learned to be sorry. The question is whether they learned it from us or whether they just noticed what worked.

More on the AI revolution as it unfolds awkwardly across London at The London Prat. Further sincerity-optional commentary at The Onion and NewsThump.

SOURCE: https://prat.uk/ai-now-capable-of-saying-sorry-without-meaning-it/