Would YOU Pass the Turing Test?

A month or two ago, I was summoned to the Microsoft HQ here in Prague for a conference on AI. A bit of bragging about the generative spoken language models and so on. Above-average catering and a bunch of sales hawks perching for prey. It’s a rough game since all the start-ups they represent offer the same service built on the same technology backed by Azure Cloud. I bet nobody there had ever heard of CrowdStrike back then, but that’s another story. The current AI is advanced enough to replace the workforce even in a niche Slavic language with some 11 million native speakers, flexible word order, and more irregularities than grammar rules. Moreover, it already makes business sense to localize the AI - so much for its scalability.

Source: Investopedia / Joules Garcia, via

ChatGPT-4o (and similar models) are marketed as an IQ 140 tireless mind under your thumb: get the subscription and never have to do trivial stuff again. Sure, that brilliant mind is quite autistic, and you can always find ways to corner it. And it can always go rogue – it’s called hallucinating AI, which I find hilarious. Yet AI can match real people in various fields – I’ve held several PoCs to prove that. People can also hallucinate, anyway.

Would this AI pass the Turing Test? Likely so. It can read and describe a picture, shows certain levels of creativity, is usually coherent, remembers previous conversations, and adheres to recommendations and suggestions it received. Pretty much like a human, even though it does not actually think. At least not the way old-school sci-fi authors imagined. Training AI for business purposes is akin to onboarding a new employee – except that a person needs information served bit by bit to cope with it. An AI model can digest loads of information at once.

And here’s where it becomes tricky. Would every person pass the Turing Test when facing an advanced AI model? Can genuine stupidity stand the challenge? I’ve had several sad conversations with people who could fail – I guess everyone has. Not being the strongest puppy in the basket is not a crime, and you can hardly be blamed for that, but the distinctive feature of such people is usually a lack of self-reflection, and AI is programmed to use it.

Some people are capable of repeating a single subclause they memorized (What a good boy this puppy is! Here’s your treat!). “Man, there’s always context. Law is not a quoting competition. At least in democratic countries.” An AI starts mimicking thinking and usually takes the hint, correcting itself. Many people won’t; it’s simply beyond them. Sad but true. Unlike me, an AI model would have the patience to talk to such people. I’ve given up.

When a strong puppy runs into an AI troll ;)

Yet those with little or no self-reflection are exposed to AI in another Turing-like competition. And it seems to be a lost cause. They are actively being targeted by robotic troll farms. We are all aware they exist, not only the infamous Russian ones. Just like an AI assistant can attend to all customers in no time, handling their requests and inquiries just like you trained it, it can also spread bullshit 24/7. And it is already doing so. I am sure that AI analyzes the best targets based on their social media content and then strikes.

How could we fight this lost battle, as humans lack the never-ending patience machines have? With another AI! Let AI Elves fight the AI Trolls! It’s gonna be rough, and there will be dead puppies all over the virtual realms – many are beyond saving, lacking even the elementary level of self-reflection, so you can prove them wrong, yet they will still insist on their mantra. Be it the single subclause they know; be it the praising of junk food with palm oil subsiding high-quality oils; be it their false impression of what tradition is… And I believe those would fail the Turing Test, being less human than the current AI models.

Let’s hope there are plenty of strong puppies out there. Anyway, would you pass?

This happens to be my #julyinleo day 25 entry. Feel free to join the challenge with own genuine long posts!

Posted Using InLeo Alpha