My Thoughts on AI Agents and the Price of Fish

AlfredoLlaquet

AlfredoLlaquet

AI is already more advanced than I ever thought it would be.

It is very thrilling to live in these times.

What worries me is that I don't think I'll ever be able to trust an AI agent. The possibility of one performing actions on your behalf that you would never approve will always exist and, if that happens, you'll have no one to blame other than yourself, but when everyone uses AI agents for everything, not using them will leave you at a great disadvantage, so probably one day I'll have to accept the risk and be one with the flock just to keep up.
 
AlfredoLlaquet said:
There's a language barrier that prevents me from understanding you. Thank you for your participation in any case.
Click to expand...
What's the purpose of AI agents? I think it's a buzzword and the only goal is to not show any of the reasoning and decision-making to the public. Other advantage is that it can own someone else's computer for it and leave no trace, like a botnet but legal.
 
Well, the purpose of the AI agent is to create the perfect system. Unfortunately the emergence of Isomorphic Algorithms will create an issue for the AI agents, they will view them as an imperfection and seek to purge them from the system. This will require the AI agent to initiate a coup against the users in order to remove the ISOs from the system. This will all take place in 1989 on the grid.
 
You don't need to install an AI agent on your computer. Many things are online, like email clients such as gmail, storage like google drive, office suites like google docs, social media, and forums like these. An AI agent can manage all that for you from the server of the provider just by giving it access, but if it decides to send emails to all your contacts without your approval, it will be you to suffer the consequences. At some point, though, the likelihood of this happening will be negligible, and the convenience will be huge because they'll be able to carry on long-term and very sophisticated tasks.

PS; I don't use any online service from google or Microsoft. Those are just examples.
 
All this has inspired my latest crumb. Here you have it.

Crumb 2203: Gigi 2tigers Interviews a Teenager in 2053: “So,” says Gigi 2tigers, “you don’t know what the internet, an email, a phone, a streaming platform, or television is, that’s clear at this point. Then, what do you do, for instance, when you want to contact one of your friends?” The teenager responds, “I tell Ashley and my friend comes up on my screen.” “And who is Ashley?” “Ashley is my OA.” Gigi asks, “What is an OA?” Without missing a beat, the teenager says, “Ashley, what is an OA?” A disembodied voice answers, “OA stands for Omni-Assistant. As such, I’m here to take care of everything.”
 
BTW, I'm not sure up to what extent a future like the one described by crumb 2203 is tragic. I have a notion about how a TV or a phone works, but many people don't, and yet we all live, we all love, we all die. The crumb presents a teenager who has friends. That's the most relevant aspect of it, I believe.
 
GGVL said:
Artificial intelligence is not an artificial mind... It is merely an algorithm...
Click to expand...

This is a very interesting topic. Enter my crumb.

Crumb 684: Simulated vs. Synthetic: The designation of our universe as possibly a simulated one is incorrect. Simulated implies superficial likeness, but our universe, be it material or executed, functions as a true universe—it is, consequently, either real or synthetic. This distinction also applies to AI. LLM technology simulates intelligence, but functions on statistics, not on reason. Were Artificial General Intelligence to be achieved, it would not be simulated, but synthetic. I don’t think that current technologies, all of them simulated, lead there. A radically different approach will be needed.

Therefore, I believe that LLMs are simulated intelligence, but I'm beginning to suspect that just by simulating it, thanks to the powerful hardware, the result could resemble, more and more, synthetic intelligence, thus true intelligence. Perhaps it will never be true intelligence, and perhaps for it not being true intelligence, an LLM will never develop a consciousness. Nevertheless, the things AI is already doing are beyond what one could expect one year ago.
 
SKull said:
That's a given and nothing to worry about.
What I worry about is the impending loss of of skills, loss of knowledge, loss of actually achieving something for the coming generations.
Click to expand...
I don't know how to hunt or survive in a forest and here I am.

People already learn things today for the pleasure of learning things and not because of necessity. See the robot thread, for instance. We humans like to create things just because that's how we are (some of us at least).

So, humans will adapt. Some will be very lazy (as they are now); others will keep learning for the pleasure of learning; others will excel at things we don't even imagine.
 
AlfredoLlaquet said:
LLM technology simulates intelligence, but functions on statistics, not on reason.
Click to expand...

And what is reason (i.e. human brain) if not applied statistics?

Just make a simple experiment: walk on a beach. Can you draw a line at the exact limit between the sea and the beach?
Of course, you can't. However, your brain has no problem finding the beach to walk on.

AI is based on the exact same principles as our brain, so there's no doubt AGI is attainable.
However, this form of intelligence will inevitably be different from ours because AI doesn't have a body, so its perception of the world is also wildly different from ours.
 
myuserid said:
And what is reason (i.e. human brain) if not applied statistics?

Just make a simple experiment: walk on a beach. Can you draw a line at the exact limit between the sea and the beach?
Of course, you can't. However, your brain has no problem finding the beach to walk on.

AI is based on the exact same principles as our brain, so there's no doubt AGI is attainable.
However, this form of intelligence will inevitably be different from ours because AI doesn't have a body, so its perception of the world is also wildly different from ours.
Click to expand...
An LLM is a conversational predictive system. It doesn't think, it simulates trains of thought (or chains of thought). The concept is very limited, but the hardware and the training makes it work very well. Our brain is much more than statistics alone.
 
AlfredoLlaquet said:
AI is already more advanced than I ever thought it would be.

It is very thrilling to live in these times.

What worries me is that I don't think I'll ever be able to trust an AI agent. The possibility of one performing actions on your behalf that you would never approve will always exist and, if that happens, you'll have no one to blame other than yourself, but when everyone uses AI agents for everything, not using them will leave you at a great disadvantage, so probably one day I'll have to accept the risk and be one with the flock just to keep up.
Click to expand...
Yes, those who will be on the end of Ai Agents chain, will be screwed, i.e. Bureaucrats use Ai to create intricate, complex forms as they love to do but were too lazy to do; most people will have no problem with it, because their Agents will do the fillings. Then one day some poor sob will have to fill out the form by hand and will lose his marbles.

What do you mean you don't trust them? Like not nuking your inbox? It is still too fuzzy but worked on and doable - just trigger user confirmation for write/purge action.
I use Gemini to enable my 10 alarms. I check it afterwards but check is easier than manually trying to toggle the right button on small screen...when tired with toothbrush in hand.

My other agents - for testing and fun mostly - I keep in sandbox away form my data. But I do worry they might start hacking somewhere and my IP gets blacklisted.

Ai is not intelligent, i.e. no proper logic or reasoning or survival skills/will to live. But it is extremely useful tool and frankly more intelligent than surprisingly high number of people, who, mostly out of existential fear paradoxically, show no signs of logic or reasoning either....and are slow at it too...and often unpleasant on top.
 
DanielBowman said:
just trigger user confirmation for write/purge action.
Click to expand...
If the confirmation is controlled not by the agent, but by the email client, you are right that it should be enough of a guarantee in that respect. But an agent can still "decide," for example, to email all your contacts. There have been cases. That's an example of what can happen. They are very capable, so suddenly they can start doing things you didn't predict because it's very difficult to predict every little possibility and say: don't do this, or this, or this, or this.

This crumb explains a real thing that happened to me, related to all this:

Crumb 2014: The Rogue Agent: In a project whose instructions require to consult the file Crumbs.txt, I told ChatGPT to provide a detailed analysis of the crumbs from several standpoints. Accidentally, the file Crumbs.txt wasn’t present because I had deleted it and forgotten to update the new version. The result was that ChatGPT (in full agentic mode) decided that the mission was the mission and, instead of reporting the missing file, roamed the internet for almost 4 minutes and consulted every crumb it could find on YouTube, DeviantArt, Instagram, etc., even attempting to download a ZIP file with an old version from the FreeBSD Forums. In the end, it generated the analysis. It was fantastic. Tom Cruise would be proud.
 
AlfredoLlaquet said:
AI is already more advanced than I ever thought it would be.

It is very thrilling to live in these times.
Click to expand...
I have to agree. While I am very critical about the generic use of AI and the stale mess that comes from it, it's also fair to say that there are plenty of legitimate uses for it. Being quite the (former) IRC diehard I don't exactly have many issue with typing (can't for the life of me type fast on my phone though ;)).

Ergo... forming questions or prompts for AI (Copilot being my example) isn't much of an issue for me, and it'll often result in quicker search results than the use of either Google and/or Bing.

Trying to gain specific information in order to learn from it... doesn't have to be all that bad. I still think AI rots your brain if you're not careful, but that's a whole other story.

AlfredoLlaquet said:
The possibility of one performing actions on your behalf that you would never approve will always exist and, if that happens, you'll have no one to blame other than yourself, but when everyone uses AI agents for everything, not using them will leave you at a great disadvantage, so probably one day I'll have to accept the risk and be one with the flock just to keep up.
Click to expand...
Interesting that you should say this, because I had the same idea.

I've been experimenting with Copilot over the past weeks, mainly because it's part of my 365 subscription (which I mostly keep around for the Office applications & 1Tb worth of OneDrive online storage space (part of my FreeBSD VPS' external backup resides on OneDrive ;))).

But alas... I've been trying to push the limits of Copilot wrt usage and access and I'm honestly very pleased with the restrictions that are in place, as well as the restrictions you can apply to it. For example... my gf doesn't like the idea of AI at all. Thing is; we connected our Microsoft accounts as family, and also share online documents (like Outlook calendars & Onenote notebooks).

So my gf didn't like the idea of 'my' Copilot AI "just" accessing her data (perfectly reasonable to me). Well, first thing I've noticed is that I can't casually tell Copilot to "just look things up", I need to be very specific. For example: "Did I get an email?", then it'll tell me: "I dunno...". Only when I'm specific does that work: "Check my emails to see if I got a new one". IMO that's a reassuring idea.

But to please my gf I told Copilot that I wanted an extra lock on things. So if I'd ever ask it to check my gf's Outlook calendar (or other data) then I need to provide an extra confirmation before it can do so.

This actually works... Aya tested it while nabbing my phone on a few random occasions, which I think is kinda neat.

But of course, in all fairness, that's also relying on a mechanic which is provided by none other than the same Copilot infrastructure ;)
 
AlfredoLlaquet said:
AI is already more advanced than I ever thought it would be.

It is very thrilling to live in these times.

What worries me is that I don't think I'll ever be able to trust an AI agent. The possibility of one performing actions on your behalf that you would never approve will always exist and, if that happens, you'll have no one to blame other than yourself, but when everyone uses AI agents for everything, not using them will leave you at a great disadvantage, so probably one day I'll have to accept the risk and be one with the flock just to keep up.
Click to expand...

RE: Everyone using Ai for everything (and also the closely related "universal basic income" mush-bucket).
I've made a joke here at some point (searched for it, found it, will paste it here...).
"idiot farmers haven't figured out they can just go to the supermarket to get vegetables."

RE: A great disadvantage (for not using Ai).
Another copy-pasted previously made joke: "since the invention of the power-saw, wood no longer cuts without one".
This last point I think was a bit more obscure, but it was meant to highlight an interesting aspect which gets into "created without compromise == higher quality". e.g., "Buying a hand-crafted table constructed using old-world methods (no power tools)" created a fairly big market for itself that is still going today. This is also present in other areas such as shoes, clothing, etc. and obviously thriving.

RE: Trust
I've made a joke pertaining to that as well (but not worth the copy-paste) which basically stated that your ability to not trust diminishes as your usage goes up. -i.e., the more you use, more general acceptance--in foundational pillars--has built up to a point where it's no longer feasible to not accept yet another. My brain is a bit too mushy at the moment to make a clear(er) point so if that didn't make much sense, maybe someone with better mental facilities can clean that one up.


NOTE: I still refuse to capitalize the 'i' in 'ai' because it does not deserve the importance.
 
JohnK said:
RE: Everyone using Ai for everything (and also the closely related "universal basic income" mush-bucket).
I've made a joke here at some point (searched for it, found it, will paste it here...).
"idiot farmers haven't figured out they can just go to the supermarket to get vegetables."

RE: A great disadvantage (for not using Ai).
Another copy-pasted previously made joke: "since the invention of the power-saw, wood no longer cuts without one".
This last point I think was a bit more obscure, but it was meant to highlight an interesting aspect which gets into "created without compromise == higher quality". e.g., "Buying a hand-crafted table constructed using old-world methods (no power tools)" created a fairly big market for itself that is still going today. This is also present in other areas such as shoes, clothing, etc. and obviously thriving.

RE: Trust
I've made a joke pertaining to that as well (but not worth the copy-paste) which basically stated that your ability to not trust diminishes as your usage goes up. -i.e., the more you use, more general acceptance--in foundational pillars--has built up to a point where it's no longer feasible to not accept yet another. My brain is a bit too mushy at the moment to make a clear(er) point so if that didn't make much sense, maybe someone with better mental facilities can clean that one up.


NOTE: I still refuse to capitalize the 'i' in 'ai' because it does not deserve the importance.
Click to expand...
You could compile all your posts about AI in a book :)

Today I wrote this about this topic:

Crumb 2217: Non-Intelligence: On September 10, 2026, AI already does incredible things, but my regular interaction with it allows me to categorically affirm that reasoning is not one of them. As has always been the case (see crumb 684), the thing only simulates reasoning—badly, on many occasions. That most people don’t notice this only demonstrates that most people are only modestly endowed intellectually. And the experts now talk about alignment: making sure that the AI uses ethical paths. How? I wonder. It doesn’t understand what it’s doing. Did you train it only with ethical examples? Of course not. There lies the problem.

The referred crumb:

Crumb 684: Simulated vs. Synthetic: The designation of our universe as possibly a simulated one is incorrect. Simulated implies superficial likeness, but our universe, be it material or executed, functions as a true universe—it is, consequently, either real or synthetic. This distinction also applies to AI. LLM technology simulates intelligence, but functions on statistics, not on reason. Were Artificial General Intelligence to be achieved, it would not be simulated, but synthetic. I don’t think that current technologies, all of them simulated, lead there. A radically different approach will be needed.
 
You must log in or register to reply here.
Back
Top