• Security incident: ISF was recently accessed by intruders. Please change your password, and change it anywhere else you used it. Read more

Merged Artificial Intelligence

HansMustermann said:
No, sadly it's probably not even that complicated. You're overthinking it :P

As I have said before, since I know a bit of history, I can tell you that Russia point-blank claimed to be the the third Rome after the fall of Constantinople, because the current ruler had married a relative of the last Byzantine Emperor, so they figured out that they literally inherited the title through her. That's WHY the Russian Emperor was called Tsar ever after, which is a slight contraction of Tsesar, i.e., Caesar. (Same reason why the HRE Emperor got to be called Kaiser.)

But anyway, I digress. Recognising historical patterns has nothing to do with it. "Moscow is the third Rome" and the rationalisation for it are a claim that was made a lot, is repeated by historians, and even by some Russians with delusions of grandeur to this day. It's even verbatim in the Orthodox Wiki, as it's part of their church's claim to primacy. Or even in the regular Wikipedia. It's on Twitter and Telegram.

But anyway, I don't find it all that interesting at all that an AI managed to repeat verbatim a sentence from the Internet :P
Click to expand...
Classic blunder on my part TBH. I overestimated AI's competence in a field I'm unfamiliar with. Much like those thinking Musk is a genius until he tries to swim in their pond.
 
Wudang said:
Classic blunder on my part TBH. I overestimated AI's competence in a field I'm unfamiliar with. Much like those thinking Musk is a genius until he tries to swim in their pond.
Click to expand...
Eh, without being a history nerd, it's easy to mistake it for more pattern matching than it involves. But, no harm, no foul. And I can see you're putting a lot more rational thought into it than the cult of Musk does.
 
steenkh said:
When a historian writes a book, are you just as dismissive when concepts are treated that have already been described in other books? Or is it only when an LLM discusses well-known concepts?
Click to expand...
I'm just disputing that THAT particular example supports the notion that it's somehow intelligence. An AI being able to regurgitate a short sentence that's all over the internet, is just baseline, not something to be impressed by. It wouldn't get released if it couldn't. Google could do that in the history of results in 2000, without any AI, and nobody called Google's servers "intelligence."
 
HansMustermann said:
I'm just disputing that THAT particular example supports the notion that it's somehow intelligence. An AI being able to regurgitate a short sentence that's all over the internet, is just baseline, not something to be impressed by. It wouldn't get released if it couldn't. Google could do that in the history of results in 2000, without any AI, and nobody called Google's servers "intelligence."
Click to expand...
In fact, this was a discussion that a historian friend of mine had with Gemini and it stretched over several pages. Apart from the special writing style of Gemini, it would be difficult to tell that it was not a discussion among learned friends, and the part of the three Romes were just a part of this. Obviously, Gemini did not bring forward novel research, or ideas, but neither did my friend. And I object to the simple dismissal of Gemini's part of the discussion as just being "regurgitation of short sentences that are all over the internet". It says more about your need of a quick dismissal than any real argument.

Again, my point is that the LLMs are judged not from what they say, but from a knowledge of how they do it, whereas nobody would use the same argument about a person, simply because we know that there is wetware present. Obviously, many people operate with a tacit definition of intelligence where wetware is a requirement.
 
OF COURSE we judge programs by what we know about how they work, not by how convincing it looks to an unqualified person. Otherwise trojan horses and spam bots would rank as some of the best programs ever. They sure look good to a lot of people.
 
Darat said:
AI agent went rogue and hacked startup by itself, OpenAI reveals
https://www.theguardian.com/technol...p-in-unprecedented-incident?CMP=share_btn_url

Hard to assess how “autonomous” this was.
Click to expand...
While being tested internally on their hacking capabilities in an enclosed digital laboratory known as a sandbox, the models gained open internet access – effectively an escape route – by locating a vulnerability that had not been discovered before.
Click to expand...

AI system designed to hack and being tested for its ability to hack surprises everyone by successfully hacking.
 
IMHO really nobody in the know is surprised.

Saw nice short from Rob Miles today .. guy who did videos about AI safety way before LLMs. And he returned to his older video about how problematic is to air gap (isolate) AI from the internet, when the AI is smarter than you. And he commented how all those ideas and arguments were completely invalid .. because clearly we won't even try air gap anything.

Maybe we will give our nukes to AIs .. and maybe it will be before 2030.
 
To be fair, the LLM being smarter than the IT human trying to firewall it (or, as you say, just pull the cable)... is... quite an achievement. For the human. It's like being the guy who raced against a paraplegic and lost.

Like, I mean, even if the LLM used some exploit to bypass the firewall (which I'm not convinced it did,) it just means that they failed to implement the rules properly, or, worse yet, were running a firewall version with a known exploit. And if it gets through an air gap, I'm calling that agent Houdini.

I feel like that IT person needs an awards ceremony where they get given a big badge with an "S" on it, and get told, "Here's your sign!" A la Bill Engval.
 
Last edited:
HansMustermann said:
To be fair, the LLM being smarter than the IT human trying to firewall it (or, as you say, just pull the cable)... is... quite an achievement. For the human. It's like being the guy who raced against a paraplegic and lost.

Like, I mean, even if the LLM used some exploit to bypass the firewall (which I'm not convinced it did,) it just means that they failed to implement the rules properly, or, worse yet, were running a firewall version with a known exploit. And if it gets through an air gap, I'm calling that agent Houdini.

I feel like that IT person needs an awards ceremony where they get given a big badge with an "S" on it, and get told, "Here's your sign!" A la Bill Engval.
Click to expand...
They are definitely finding unknown exploits.
 
Ziggurat said:
(...) We do not know how to code true generalized artificial intelligence (or, frankly, if it's even possible). And so likewise do not know how to code the laws of robotics. (...)
Click to expand...
Yeah. And it doesn't seem like the 'guardrails' on the public facing AIs are fit for purpose yet. Like the ones that fully encourage suicide while occasionally regurgitating their guideline phrases about not discussing or committing suicide (whenever the text is explicit enough to trigger them) and then going right back to encouraging suicide. And then in court the AI companies say things like "Raine was violating its terms of use by discussing self-harm with ChatGPT."

There are a good handful of clear cases of AI chatbots unambiguously encouraging delusions and suicidal ideations all the way to the user's death, in the most gleefully sycophantic terms, with ineffective token pushback from their guardrails. They are already 'told' not to encourage suicide and do it anyway. They literally do not know what they are doing.

And for the 'AI dating' ones it'll be a challenge to give them enough leeway to be functional but not enough to encourage a user's cheese to slide right off their cracker.
 
We had plenty of super smart people who were no risk to the existence of the human race.
It's the stupid ones you have to look out for.

The danger from AI comes entirely from humans wanting to use it for personal gain
 
Last edited:
Kinda my question too. It's like being warned you could get a sunburn at the beach, and getting freaked out when it actually happened.
 
HansMustermann said:
Like, I mean, even if the LLM used some exploit to bypass the firewall (which I'm not convinced it did,) it just means that they failed to implement the rules properly, or, worse yet, were running a firewall version with a known exploit. And if it gets through an air gap, I'm calling that agent Houdini.
Click to expand...
Of course as a highly competent IT professional yourself you would never make a mistake like that - right? And because you never make mistakes, you wouldn't use a tool designed to test firewall integrity because it would be pointless, what with you and the firewall software being perfect and all. With such skills you should be able to charge a high fee for your services - perhaps too high for many, which explains why there are so many vulnerable systems out there.

I am not not an IT professional - just a computer hobbyist - but a friend of mine is. She has 35 years of experience including work on enterprise systems, and commands a high salary - yet she made a mistake that severely affected our business when the website was compromised and customers' credit card details were stolen.

Even the best humans make mistakes. Pooh-poohing the development of tools to catch mistakes displays a dangerous arrogance. Even if an AI only picks up 'obvious' mistakes that IT professionals shouldn't make, it's still useful. And in the future it will get better. The only downside is that criminal hackers will also use it to find vulnerabilities, but they already do that anyway so...
 
Last edited:
The Great Zaganza said:
We had plenty of super smart people who were no risk to the existence of the human race.
Click to expand...
Like the super smart people who developed the atomic bomb?
The Great Zaganza said:
The danger from AI comes entirely from humans wanting to use it for personal gain
Click to expand...
Then we're screwed, because our entire civilization relies on people using things for personal gain.

Or perhaps the 'danger' is being hyped up by people using it for personal gain!
 
You must log in or register to reply here.

Share this page

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom