Wow a web site generated using AI[1]. (or perhaps a human using AI)
Anecdotally, when I was attending college there was a 12 year old girl also attending and in some of my classes, particularly my freshman physics class. She was knocking the curve off with high scores on all of the exams. I got a chance to talk to her at lunch one day and it turned out she had an eidetic memory. It was amazing, she could tell you what was on any page of the text book perfectly. That allowed her to recall worked problems in the text that were identical in form to the question on the test, and she could then use the same steps to solve the test problem. But, and this was an important part, she didn't really understand physics. Whenever our conversation went into areas where she could have used physics principles to derive an understanding or at least a good guess at some of the depth of a new topic, she did not. That didn't hinder her progress through school but I had to believe that at some point it would.
After that experience I started paying more attention to people who "knew" facts, and people who "used" facts, which is to say that people who had learned something and understood it, would use that learning to extrapolate into new areas, open up places they didn't understand, and pursue new knowledge about those gaps. And there were people who would rebut arguments with "facts" but seemed not to grasp the fundamental principles at issue.
AI generated "answers" to prompts have exactly the same properties as answers from people who know facts but don't understand them.
I would guess that the article in question was generated with some prompts of the form, "Describe how an IMSI catcher works for each type of network." If you're a human and you read the answer and noticed that 5G was different you can add the click-bait headline and voila, article!
And yet for someone who understands how IMSI catchers work and understands the general compatibility environment of the cell phone networks, they would point out that most phones are designed to work "around the world" which means with all types of networks 2G/3G/LTE, and so even if the world around you is LTE/5G if you pop up a GSM cell tower signal a modern phone will see it and say hi. And then they would go on to describe that WiFi and Bluetooth device hardware (MAC) addresses are unique too, and those are also sent around if you bleat out your an open wifi network or a lonely bluetooth device. Finally it would point out that even with the 5G "SUCI", that value is unique to your phone and even if you don't give someone enough information to reverse map your phone to you, it is absolutely enough information to keep track of where this particular phone has been over time.
But all of that context is related to understanding why you would even want to capture and IMSI number and how the entire system was designed to make that easy even though now that is seen as a vulnerability.
So if you've spent some time recognizing the difference between people who are talking about something they understand and people who are talking about something they read about but don't understand, stuff written by AI just sort of pops out at you like that.
[1] All the generated images at the bottom was a dead giveaway but the structure of the article was also indicative of an LLM construction.
This is a very interesting comment. When I read your physics story, I thought you would be getting to the similarity to current llms. However hallucinations seem like a different issue that the young student might not have. If she incorrectly matches some scenario to a text match, maybe some hallucinations. Some humans are confident in making comments about things I don't understand, like you know who. But many humans somehow have a concept of their limited knowledge. When they add that to LLMs, that will be powerful.
I pretty much agree with this, having some way to indicate model boundaries in an LLM parameter space to create back pressure on token generation would help a lot here.
For me though the interesting bits are how the lack of understanding surfaces as artifacts in the presentation or interaction. I'm a systems person who can't help but try to fathom the underlying connections and influences that are driving the outputs of a system.
1) Yes, it's still possible to convince a phone to connect to older RAT generations. However, the idea is that as those are phased out, it's unlikely that they'll be enabled on UEs, so phones likely won't connect and "say hi" as you say. For example, 2G is already being disabled on many devices. It'll be a while before 2G-4G is fully phased out and IMSI Catchers become completely infeasible, but I think that it's safe to say that "5G got it right" in finally solving this issue.
2) I don't think you have a good understanding of how SUCIs work if you think that it's unique to a device. The UE generates a fresh ECC ephemeral public key every time it sends its SUCI (which isn't often to begin with due to GUTIs, which are one-time use and only assigned post-ciphering). You can read more about it here: https://medium.com/@aditya.koranga/ecies-in-5g-core-supi-to-...
Wow a web site generated using AI[1]. (or perhaps a human using AI)
Anecdotally, when I was attending college there was a 12 year old girl also attending and in some of my classes, particularly my freshman physics class. She was knocking the curve off with high scores on all of the exams. I got a chance to talk to her at lunch one day and it turned out she had an eidetic memory. It was amazing, she could tell you what was on any page of the text book perfectly. That allowed her to recall worked problems in the text that were identical in form to the question on the test, and she could then use the same steps to solve the test problem. But, and this was an important part, she didn't really understand physics. Whenever our conversation went into areas where she could have used physics principles to derive an understanding or at least a good guess at some of the depth of a new topic, she did not. That didn't hinder her progress through school but I had to believe that at some point it would.
After that experience I started paying more attention to people who "knew" facts, and people who "used" facts, which is to say that people who had learned something and understood it, would use that learning to extrapolate into new areas, open up places they didn't understand, and pursue new knowledge about those gaps. And there were people who would rebut arguments with "facts" but seemed not to grasp the fundamental principles at issue.
AI generated "answers" to prompts have exactly the same properties as answers from people who know facts but don't understand them.
I would guess that the article in question was generated with some prompts of the form, "Describe how an IMSI catcher works for each type of network." If you're a human and you read the answer and noticed that 5G was different you can add the click-bait headline and voila, article!
And yet for someone who understands how IMSI catchers work and understands the general compatibility environment of the cell phone networks, they would point out that most phones are designed to work "around the world" which means with all types of networks 2G/3G/LTE, and so even if the world around you is LTE/5G if you pop up a GSM cell tower signal a modern phone will see it and say hi. And then they would go on to describe that WiFi and Bluetooth device hardware (MAC) addresses are unique too, and those are also sent around if you bleat out your an open wifi network or a lonely bluetooth device. Finally it would point out that even with the 5G "SUCI", that value is unique to your phone and even if you don't give someone enough information to reverse map your phone to you, it is absolutely enough information to keep track of where this particular phone has been over time.
But all of that context is related to understanding why you would even want to capture and IMSI number and how the entire system was designed to make that easy even though now that is seen as a vulnerability.
So if you've spent some time recognizing the difference between people who are talking about something they understand and people who are talking about something they read about but don't understand, stuff written by AI just sort of pops out at you like that.
[1] All the generated images at the bottom was a dead giveaway but the structure of the article was also indicative of an LLM construction.