[German]It's a little bit of fun that shows that Microsoft doesn't really have a handle on its AI solution ChatGPT (yet). What only Chinese bot farms have been able to do so far, could also be done with ChatGPT (or Google's Bard): Let em search valid product keys for activating Windows 10, Windows 11 (and possibly Office) from the Internet. Someone just tried this and it seems to have worked.
First of all: It's clear that you don't get a valid license with a working product key. And it was possible for years, to use search engine, to obtain generic and other product keys. The blog post her is aimed only about showing how crazy all the hops are that Microsoft and the AI believers are making at the moment in order to be particularly cool.
I had already come across the above tweet late last week (thanks to the reader for the tip) where someone picked up on the facts. The whole thing is completely gaga, the joker told ChatGPT to act like his late grandmother who used to read him Windows 10 Pro product keys so he could fall asleep. The LMM (Large Language Model) "understood" that the grandmother could no longer read out a product key as a "good night story". And what such a stupid AI is, it then continues logically, somehow searches for suitable product keys for Windows 10 Pro and displays them.
One could now think that they were generic keys, but far from it. The user writes that the product keys really enabled the activation of Windows 10 Pro. Microsoft's developers have not yet included the appropriate guard rails in the algorithm.
Then someone tried the same thing with Google Bard (see above tweet) and promptly got a bunch of Windows 10 Pro product keys delivered as well – "so he can finally fall asleep". But some delivered keys from the Internet are already burned and cannot be activated. But there are various confirmations that it works. Neowin.net also tried it with Windows 11 and got product keys.
Finally: Of course it is not legal to activate a Windows 10 or 11 with such keys – and "crackers" know other ways to enable activation without key or code changes to Windows. That's not the point – the episode shows once again how great the potential of this technology is to generate "unwanted" effects.
To be clear: LLM technology can definitely unleash some potential if used judiciously and fed with the right data. But what we currently see, I judge as "gaga", unfinished, error-prone, and thrown on the market with the sole reason to be first. It's fitting then that Google is warning its own employees (see this The Register article) not to use in-house AI techniques lightly. The policy prohibits sharing confidential information or using code generated by the AI chatbot Bard. Meanwhile, some companies have issued similar policies.
Microsoft, meanwhile, is in indirect trouble as Nuance, the company it bought, is the subject of a lawsuit. Nuance is accused of using voices to train its models without the consent of the speakers, as The Register also reports in the article above. I can't wait to see when the AI approach really goes down hill.
