On the lighter side

We all have that at times, the setting of a lighter side. You know the average romcom with sex, a little violence, the setting of blackmail and of course, some piece of software. The average day in the life of an fake AI. So I was ‘rudely’ awakened by news (at https://www.bbc.com/news/articles/cwyklykn5dwo) where we see ‘Anthropic accuses Chinese rival Alibaba of illicitly extracting AI capabilities’. So it made me laugh, b because this comes at the near start of a bubble heading straight for the HQ of Anthropic. Now let me show you what I was having in mind with all this and best I use graphics for that setting.

So we see the application on the left, the data on the right and in the middle we get the setting of that fake AI, you see it has DML and LLM, which I represented as a separate stage, but it could be one big thing, the coding is in the middle and there is interaction between the three like any application would have. So the middle part could be part of the application (it likely is), but for the clarity I wanted to show it like this. Because the picture fits better for the explanation. So the question becomes (the sound of dramatic horns in my mind)

“US artificial intelligence (AI) giant Anthropic has accused Chinese e-commerce and technology firm Alibaba of “brazenly” and “illicitly” extracting its Claude AI model’s capabilities.” How is this possible? I am not saying that it cannot be true, because that requires evidence, but if we see these parts, how blazingly stupid is anthropic to let someone else have a go at this. Beside this, what EXACTLY is “extracting its Claude AI model’s capabilities”? You see, when you see the image, the capabilities are shown in the application and cannot proceed without data (or less likely so), so as we are in the bubble setting this so called move sounds like a joke and with the added “In a letter sent to two members of the US Congress, the San Francisco-based company said operators linked to Alibaba carried out almost 29 million exchanges with Claude using thousands of fraudulent accounts in what it called the largest extraction campaign of its kind.” It seems like there is a massive security lack in all this (that is, if there is a transgression stage). But the setting that we see with “operators linked to Alibaba carried out almost 29 million exchanges” so as I see it, in 2014 we had the Cambridge Analytica scandal, where Facebook got ‘relieved’ of a whole lot of data. Doesn’t anyone learn from that experience, as such we get a repetition of all this? But I hope the story is clear. How was this even possible? As I see it “According to Anthropic, the campaign was carried out through what are known as “distillation attacks”, which extracted answers from a stronger AI model to train a weaker one.” This is a debatable setting (not stating it cannot happen), but the image I ‘created’ shows that a distillation attack requires a lot of information that requires insider knowledge to be successful. 

As such, I am not saying that Alibaba and the 40 fighters for the Palestinian cause is innocent, but I have doubts on the entire setting. I personally see this as a Dutch SNS setting. Where the massive mortgage (the invested dollars in Anthropic) are written of by putting it into a bad bank and letting that bad bank collapse. As such you need to be aware that I could be wrong, it is based on expected behavior and speculative settings, so do not take my word on it, but consider that at present the BBC is spinning you a yarn by presenting the data from others. Just so you know.

So, this is how I got my 05:00 wake up call, thank you BBC. Now it is time to get some coffee and optionally have breakfast afterwards.