A thread for unboxing AI

AI drones is just the latest variant of the projectile weapon and indirect fire. I think it is a dubious argument that it is somehow going to make war war worse or deadlier than artillery shelling, mortars, aerial bombardment and missile strikes.

Sure, it sucks to be a soldier waiting for the buzzing sound of imminent death or being hunted by a robot, but I'm hard pressed to see how it is worse than waiting for a 2000 pound Mark 84 to the noggin via CAP, driving over an IED, listening to the drone of incoming carpet bombing, an artillery shell, a musket shot to the chest or being starved to death in a siege.

As for the ethical component of not using human decision making, we have done that for hundreds of years with victim-activated mines. And AI drones are not any more unethical than dropping mines over populated regions, because you do not have human control over the mine's detonation either, and you know full well that most of the casualties will be civilian.

There are good reasons to want to limit drone and drone warfare, but they are not particularly different from the reasons to limit many other types of weapons - and is a mistake to argue that it must be done because wars will become worse. Drones and AI drones are bad because war is pretty ****.

The objective of war isn't to kill soldiers per se; its to destroy the enemies combat resources. As AI and robotics become so advanced that a soldier has zero chance against them, I'm thinking the battlefield will reduce to robot vs robot, drone vs drone, etc.

per se, ergo, concordantly

The bolded seems like a considerable oversimplification and more wrong than right in some cases.

"War is God." Judge Holden, Blood Meridian.

I'm sure that you can find a million differents quotes about why people wage war. Those quotes don't change the reality that the primary purpose of waging war isn't always to destroy the enemy's combat resources. Sometimes other things -- for example, rallying support domestically -- are primary goals from the perspective of the people who start the war.

Sure but that's way outside the changing battlefield scope I was addressing.

Or they just want all the inhabitants dead so they can have the territory.

My job currently has me connecting ChatGPT to Zabbix for AIOps.

Zabbix has an official MCP server. 281 tools, fully worked out. It SUCKS for chat. You can't ask it any question and get and data back. If you do it's ten minutes making API calls. ALL the available MCP servers for Zabbix are just raw API.

I built one last week in one day with Claude in NodeJS that actually works. So I'm improving that and making it MCP compliant. So looks like I'm on the cutting edge of Model Context Protocol.

I think I need to ask chatgpt to explain what you just said.

opus 4.7 is much smarter than 4.6, if it's work doesn't pass qa review then it's figured out the best way to changed "BLOCKED" into "PASS" is to just edit the QA doc i feed it into positives with a "PASS" verdict

Laziness, hubris, and impatience are the three great virtues of a programmer. I learned this from Larry Wall's Programming Perl.

humanity was fun while it lasted

"How should I do that?"

Basic: You can't do that.
Pascal: You can do this instead.
Java: This did not work, so instead I did that
C: You could do it like that, but the right way is this.
C++: I concur, but only if you do it like that.
Perl: If you can't do that, there is always this.
Python: There is no this, only that.
C#: // Function for doing that like this.

prolog: do(?, that)

We agree on the main point which is that avoiding war is the key objective. I wasn't making an ethical or restriction on weapons case because while they matter, in the end aren't worth a hill of beans if we don't restore international order.

However I don't agree that dumb weapons out of human control like mines are in the same category as AI controlled weapons. It's like saying that nukes don't matter because we already had dynamite. The ethics and potential are totally different. Unlike with nukes there is no chance whatsoever of countries not using AI if we continue with wars lie we are. That is the point of my post. We need a renewed international effort like the UN in 1945 or we are likely to face hell on earth.

I had a few beers with a buddy of mine last night. He's manager of a small IT shop for a company that does, well, I have no clue. But, he's been with them as either a contractor or now employee for at least 15 years (maybe longer?). He started out there because he knew the guy that owned it and they needed some programming/database work done so he did it as a contractor. The guy liked his work so he continued there and has written most of the software the company uses. My buddy admits it's pretty crude stuff. I think he told me it uses access. Anyways, as with most legacy stuff written that long ago, there's a ton of spaghetti code and horrible database design. Again, he readily admits this.

Anyways, he hired a guy that told him they should be using Claude. My buddy figured why not, so they paid for it and he started messing with it. It took him a while to figure out what to ask/tell Claude, but he worked at it. He told me he finally got it to do what he wanted. It redesigned their entire database and re-wrote every spec of code. I think he told me it took a few weeks, but most of the time again was figuring out how to ask the right questions. He told me his QA department is freaking out because they have no idea how to test the volume of stuff coming at them. I told him, use ai to do that, and he replied, "already on it."

It's pretty crazy stuff. I'm glad I retired when I did. My old job is pretty much obsolete now.

The problem is that an organization with tons of spaghetti code and legacy horrors is probably also an organization ill equipped to deal with the kind of QA that vibe coding a rewrite of your entire digital infrastructure requires.

Relying on AI in both ends is not really a good answer.

We don't really know much about the effects of drones on the totality of war. We have indicators that the ratio of military casualties of death to injured increase. In large part because evacuation becomes very difficult, as loitering drones are very good at trapping soldiers in positions.

The Russia-Ukraine war is not really a good indicator. It is horrific on the civilian populace, but so was Syria where the Russians use more traditional weapons. The problem isn't the weapons, as much as it is a doctrine of purposefully destroying civilian infrastructure and killing civilians on a massive scale.

Nor is the current Iran war a good indicator of whether things will be better or worse. We have Iran lobbing drones at civilian targets as well as near them, but you also have the US firing Tomahawks at urban targets and Israel shelling cities.

So that drones make war worse overall is still difficult to gauge. Discussing which has the worst potential for overall harm is a debate which is not seeing the forest for the trees. They can all be pretty damn shitty.

Drones as in loitering munitions (not all drones are) actually have the potential to be more precise than traditional weapons used for similar targets. I despise the term "surgical strike", because there is no such thing as a surgical strike. Not in precision, civilians suffer in war and surgical strike is a politician's term to pretend they do not. Nor in intention, surgery is something you use to save lives, not take them. Still, awful terminology aside, modern drones could better at this type of attack than older technology. It depends on the operator, their doctrine and their culture.

My prediction: It will be just as shitty, just different.

What's happened so far with drones is almost nothing. If you think this isn't just the most minimal tip of the AI/robotics capabilities that are coming then we profoundly disagree.

Where we probably agree - I think we should do our best not to find out how shitty it will be. That wont be by stopping development.

Using available evidence to see where things are headed is the prudent thing to do.

Where drone warfare will ultimately take us is unknown, and it is tiring to listen to people who think they do.

Waiting until it is has been really shitty to conclude it will be really shitty is not being prudent.

There's plenty of evidence of what capabilities are coming. It is not remotely prudence to demand evidence of it having gone horribly wrong before doing anything.

What we want is people to say we were wasting our time doing something about it because it never happened.

Well, thankfully I have done none of these things, so I suggest you continue the debate with the relevant party.

Ok so we agree the evidence of drones so far shouldn't be used to argue against the risks from the AI/Robotic capabilities that are coming.

I don't get your prudence point then. It is prudent to seriously consider just how horrific the potential is from what is coming.