09:48:49
##forth
<veltas>
tpnix: Your opinion is worth infinitely more than 'AI' though :P
09:50:47
##forth
<tpnix>
veltas, I just wanted a 'second opinion' :)
09:52:20
##forth
<tpnix>
Im still busy tidying up the doc and completing the example program and didnt want to be detoured into trying to quantify any code savings with FURS
10:11:16
##forth
<tpnix>
veltas, in a way, I'm lucky with FURS as I designed all the important parts years before AI was even released, so I didnt have the temptation to cut corners back then
10:12:09
##forth
<tpnix>
veltas, and I'd never get AI to write my Forth code as that would be like buying apple pie and cream then getting someone else to eat it!
10:13:10
##forth
<tpnix>
veltas, I love Forth coding, I love the mental challenges of Stack juggling, it's a pleasure for me
10:46:26
##forth
<veltas>
I mostly use AI as a reference or way to do queries that are hard with Google
10:46:40
##forth
<veltas>
'AI' because it's not AI
10:47:15
##forth
<veltas>
I've been using LLMs for help with design questions like colours etc around my house, because I have no knowledge at all
10:47:23
##forth
<veltas>
And it seems like an expert when you don't know anything
10:47:30
##forth
<veltas>
But when you do you can see the gaps
10:57:58
##forth
<tpnix>
veltas, yes, it's easy to expect too much from LLM's
11:05:12
##forth
<veltas>
I'm still learning what they're good at and what they're bad at
11:11:50
##forth
<tpnix>
same here, Ive found that a very tight specification with lots of information helps a lot to get what I want
12:58:06
##forth
<KipIngram>
Oh yeah - AI is just full of holes and gaffs. It amazes me how thoroughly wrong it can be sometimes, but it will offer that to you with utter confidence and assurance.
12:58:50
##forth
<KipIngram>
I think it's a fine tool for helping you think and organize your own capabilities, but it is just dangerous as hell when someone tries to use it to "push beyond their envelope."
12:59:24
##forth
<KipIngram>
I worry that that's exactly what a lot of folks are trying to do with it. Trying to use it to get an "extra leg up" in the world.
12:59:41
##forth
<KipIngram>
To get the goodie without investing the work.
13:00:10
##forth
<KipIngram>
Like copying homework in high school.
13:00:33
##forth
<KipIngram>
I use it primarily as a sounding board to bounce ideas around on.
13:01:50
##forth
<KipIngram>
Mostly for hobby stuff (it's kind of all hobby stuff now I guess). It will invariably offer to implement things for me, in spite of me telling it over and over that that's not what I want from it. The whole point of this stuff is to do it myself in the end.
13:02:31
##forth
<KipIngram>
But "Would you like for me to do <blah> <blah> <blah>" seems to be DEEPLY embedded in its operation.
13:03:35
##forth
<KipIngram>
So much so I can't get it to stop.
13:05:49
##forth
<KipIngram>
Being super "stern" with it doesn't help - it promises to do better and then goes right on doing exactly the same thing.
13:06:44
##forth
<KipIngram>
If it were a person behaving that way I'd wind up mad as hell at him or her, but there's little point going down THAT road.
13:13:27
##forth
<veltas>
I don't worry at all, the exact same situation happened with e.g. Wikipedia, Google, etc. Some people want to cheat and some want to learn, ChatGPT etc can be valuable for the curious intelligent kids who actually want to learn
13:14:36
##forth
<veltas>
My main concern is about the centralisation of this stuff, but that's not new
13:38:02
##forth
<ajhidd>
veltas: MY main concern is that governments want/try to integrate it in state management/security. No one (I know of) tried to rule a country asking Wiki or Google for advice :)
13:40:07
##forth
<veltas>
AWS is integrated into a lot of govs
13:41:40
##forth
<veltas>
It's a lose-lose situation
13:58:00
##forth
<KipIngram>
ajhidd: Good point.
13:58:29
##forth
<KipIngram>
veltas: Yeah, I think these "points of centralization" in our infrastructure are bad juju. They're points of failure and also points of attack.
13:59:00
##forth
<KipIngram>
But all it takes is for one to become the "most economical" way of doing something, and the world just piles onto it like bees on honey.
13:59:41
##forth
<KipIngram>
And I guess the last few days have shown us how "wired in" AWS is.
14:27:40
##forth
<veltas>
It's funny because the internet initially was all very distributed, I'm guessing for a number of reasons including resistance to nuclear attacks
14:27:57
##forth
<veltas>
And the infrastructure is still mostly quite distributed
14:28:32
##forth
<veltas>
But yeah it's cheaper to just not bother with all that and so one badly configured DNS can bring down a big chunk of tech
15:13:45
##forth
<KipIngram>
It wasn't really initially intended for use by a huge "mutually distrusting" user space. We built that onto it anyway.
15:14:28
##forth
<KipIngram>
For the purposes we want to put the network to today it really should have been custom designed from the ground up.
15:15:37
##forth
<KipIngram>
Of course part of the problem was that at the time it all started our hardware capabilities were meager by today's standards. It was almost required to focus on performance just to get it to work well enough to use at all.
15:15:55
##forth
<KipIngram>
This "throw cycles at it" thing is more recent.
15:18:17
##forth
<KipIngram>
Even if we'd known all the cute tricks we know today, we wouldn't have been able to afford to implement them.
15:19:08
##forth
<KipIngram>
We've come a long way, but we've brought an awful lot of baggage with us that seems impossible to jettison now.
15:20:17
##forth
<KipIngram>
Looking back it's pretty easy to see how it was that "profit" mentality that drove us right down this path.
16:26:42
##forth
<cleobuline>
these bot attacks on the ssh port are crazy
16:27:44
##forth
<veltas>
Don't use the default port
16:33:52
##forth
<cleobuline>
i do a deny on the pssword login
16:34:17
##forth
<cleobuline>
and root login too
16:35:07
##forth
<veltas>
Better to just refuse connection
16:35:17
##forth
<cleobuline>
yes i do it
16:35:19
##forth
<veltas>
Smaller logs