Recent content by jack action

I think the scenario would be more like I hack Harvard's website (good reputation) and hide my ToE on some webpage (say, hidden in a HTML comment, as suggested). Imagine if I can even spread my ToE that way with many websites. Then, ChatGPT finds the text and starts sharing it with anyone asking...

This is where I don't understand how it is possible to do such validation. Referring to the quote in my previous post, we are talking about "propagating misinformation", "overriding safety rules" (are the validators safety rules not included?), or "HTML hidden elements" (those might be easier to...

But, from your link, it is not successful: You are supposed to use AI as an assistant, not as an expert: Without the AI work reviewed by experts, without a "responsible and ethical use of digital technologies", it is as successful as a professor who would ask their assistant to pick and read...

I'm very curious about how AI can determine the "user's goal". How does a developer can assure safety? "AI is doing it, I trust it will do a good job"? To make sure everyone is on the same page, this is what indirect prompt injection looks like: As one can see, the possibilities are endless...

Yes, but what are you looking for in the sanitizing process? SQL has a particular grammar that must be respected, but how many different ways can a prompt be written for LLM and give the same result?

May I refer you to the thread title?

From the link: Some developers had problems dealing with SQL injection; I can't imagine the complexity of dealing with indirect prompt injection.

If I may chime in: I don't bellieve it.

But since LLMs take their information from the web, if there are no questions nor answers there anymore, how will they be able to answer questions? That will be interesting in a few years, when there will be questions about new technologies, processes, protocols, etc.

But wouldn't that change of reading NOT include the volume of the antlers already submerged without pushing them down? For example, what if the antlers sink completely below the surface (on the bottom or not), on their own? Then the reading is just the combined weight of the tank and the antlers.

@PeroK , what would you choose given there was a puzzle of some sort and your life depended on getting the right answer, and you have the choice of consulting Wikipedia, Google search engine, or ChatGPT? That is a more interesting choice especially knowing that both Google and ChatGPT have...

Knowing I'm no expert on prompt generation, let's see what GPT-5 mini from DuckDuckGo has to say: I asked for the short prompt and got: I then enter that prompt into the free image generator deepimg.io and got these from Nano Banana Free (no login needed, defined as "Supports multi-image...

If Picasso never existed and someone asked an LLM to draw the portrait of a person and the result was this: Would it be considered a bug or a feature?

$$F_b = W_{act} - W_{app}$$ $$F_b = \rho_{fluid}gV_{fluid}$$ And if the object is completely submerged: $$V_{obj} = V_{fluid}$$ Thus: $$V_{obj} = \frac{W_{act} - W_{app}}{\rho_{fluid}g}$$

With AI, LLM in particular, the ingenuity of humans is just transferred to making good prompts. For example: Also, look into AI prompt engineering: