Yesterday I talked about Facebook using automated AI system for its security system, and honestly this isn't just Facebook, Instagram and YouTube does this too, and the system is programmed to act in a unique way without reference to unique cases.
For example, there are times when I enter my Facebook password and the system flags it as "wrong" even though it's actually correct, and just trying to enter it repeatedly since you know it's correct actually makes the algo flags the login attempts as "suspicious".
Even though the mistake or error is from the AI, the AI system isn't smart enough to know that the error is from it, it automatically believes it cannot be wrong and the issue must've come from your end.
Now when the AI system fails and to you want a human review because AI cannot reason, then you fall into a roadblock because there are honestly no humans assigned the role of helping to solve unique account problems, because AI has probably taken the bulk of it.
Laziness & the tendency to be latent
Inasmuch as you think facebook has hundreds of millions of people if not billions and they cannot employ enough people to respond when their AI system fails.
I still think the system is just too daft to understand security and we cannot honestly assign the bulk of security of AI if they're not going to be smart enough to understand unique situations.
They can assign roles without a lot of peculiarities, but I understand don't even know if they want to do that. They already have a billion user and scaled 100%, so why would they do it?
This is the problem with AI at the moment.
It is currently rigid because it is programmed to be so, even with years of constant training and prompting, AI can be very inconsistent and rigid with solving real time problems because it always treats issues and situations as the same.
Now when it comes to output AI is still very much limited because it cannot reason like humans. While its problem solving skills are immaculate, it can still feel drab, sloppy and slurry, and if you're a constant user of maybe Claude, either for extensive research writing or something like that...
you'll see the inconsistencies in its output, how it tends to use similar phrases, ideas and sentences and how that it cannot stand alone unless you heavily edit it or add a human touch to it.
This is probably one of the biggest reasons why people hate AI.
it's not really because they hate it, they just hate how people use it. I'm a huge fan of AI, and how it can consistently make you better in your craft, and in all my usage of AI I've made sure to always use it in ways that complements me.
However, I think the massive dislike for AI is that people don't use it well, they depend solely on it, or they they don't using models that are more superior or add human touches where it is necessary.
A lot of companies have banned their employees from using it, and I think this is because they feel that they're paying for humans and not AI, and they could easily buy higher AI models if that's what they want.
However another thing we have to know is that AI is already smart, but it still fails in areas of peculiarities and specialities and that's why it's easier to recognize AI output and people can go "eewwww" that's AI". Perhaps in the future, when this is better, then maybe no one will have this issue anymore.
I think people are still trying to find the proper balance using AI. Most LLMs are going to begin embedding invisible statistical watermarks in the text of AI-created content at the code level to allow dissuade people from claiming AI-created content as their own. I think this will be a good thing and will force content creators to start using their brains again and consumers of content will be able to tell what's AI created and what isn't. Here in the US there's a growing resistance to AI and data centers, especially amongst younger people. However, I believe AI is here to stay and whichever country masters it first will have a pretty large advantage over the rest of the world.
AI is definitely here to stay. You're right, a lot of AI may not have visible watermarks, but as you've said, the systems are definitely going to be putting in their code watermarks in the backend so that other systems can recognize that it's AI. I think there's honestly nothing wrong with integrating AI into a workflow, I just think people will need to be more creative about it to get the best results.
Big tech companies currently using it for security are just being too lazy and not integrating people to do delicate things as security because they just don't care.
AI is good, but I think there are some people who are just using AI to do their job for them, and not putting much effort themselves. After all, they can pump out things like crazy. At this point, I think that AI isn't perfect, and its easy for it to make mistakes. So there isn't much we can say, but when it does improve, it will be hard to tell what is AI versus what is not.
Yeah, I think the heavy dependency on it is where the problem actually comes in. A lot of people would like to see AI output that looks like someone put effort into it. I think it's crazy to know that AI is not perfect, but some people are still not willing to put in the work for make the output much more better.
To be truthful about it AI is not really "intelligent" as it doesn't have its own mind for now. What it has is a collective mind of all the developers that put their mark in the code or documented that somewhere, which the "AI" ingested that information. Thus, behind the scene basically is a best match algo which sometimes get's it wrong and than enters same loop over and over again.
You're right AI at the moment doesn't really have a mind of its own, it probably would still take years of constant training to be able to do so. We probably need to to still continue tweaking the systems to deliver usable results, because the output is currently not good enough.
AI is great is my opinion, some chat bots are better than other of course. I love the pictures you can generate. Some people hate it because they still have this Terminator fear in their heads, or think it will replace humans. None of those really will ever happen. It just makes things more efficient.
It may be smart, but like you said those gaps in some areas make it unreliable at best. You have to to double check any source it ever provides. Maybe some day that will improve, I hope so.
How have you been feeling?
!PAXK