What the article finds worrisome about this is that it is a “hobbyist in a garage” building what’s essential a voice activated and commanded killing machine. They are concerned because it is not “a government lab” or similar. Nobody should be doing this, but okay.
This cat is thoroughly out of the bag. People have been building computer vision powered automatic sentry guns that shoot paint balls or nerf darts or spray water for a while now.
See, for example, this video from 2012, which was probably about one of the first ones.
In comparison, here’s a new video from a few weeks ago as part of “YouTube Maker Secret Santa” where the guy built a very good working Nerf replica of a TF2 level 2 sentry gun. He didn’t even bother talking much about the motion tracking part because it was already a solved problem (see timestamp 17:50), so instead he spent the video talking about the ammo chain design and the aesthetics (and playing with the Secret Santa gift he had received from another maker).
That’s the state of automatic AI turrets in 2025: trivial enough to omit discussion of the tracking tech and cheap enough to build to fit in a Secret Santa gift budget.
I saw a video before 2008 where a guy hooked an air gun up to a laptop and it was targeting and calculating projectile velocity and leading targets.
Removed by mod
Does it connect to HomeAssistant?
The words “I’ve been excommunicated!” or “I’m a blood donor!” they’ll never talk to you again!
Why don’t they like blood donors
It is against their religion to accept blood transfusions, among other ridiculous stuff like not being allowed to celebrate birthdays. Why though? I have no idea, it’s religion.
Nobody should be doing this, but okay.
I have to build one of these in my garage before the Chinese do it.
Maybe lock your garage so the Chinese can’t get in?
I think it’s worrisome because if hobbyist can do this, imagine what professionals with more time and money can do
He’s not just any hobbyist. He’s a professional. A literal engineer according to the article.
I mean, you could probably just do this with openCV for aiming and Kaldi or Julius for voice commands.
Why is an LLM needed for this at all?
Yeah, it’s not particularly interesting or impressive. We’ve been able to do this for years, just in the news cause it plays into a lot of people’s fears about AI.
GatGPT.
I misread the headline and thought it was saying the guy was some sort of militant centrist. 🤣
The militant centrist symbol is Jon Stewart wearing a Che Guevara hat. The T-shirt is only available in extra medium.
Gavin’ll be happy about the size, at least.
What’s the point of it? “Hi ChatGpt, the Russian enemy is like 10 degrees on the left and a bit further down from the currently aimed position. Maybe 12 degrees… Just shoot 100 times and vary the x and y degree by a radom value between 0 and 5 degrees with steps of 0.1 degrees each time based on a secure RNG, so you maybe hit the enemy. If you hit, I’m gonna tell you and give you further instructions. If not, repeat in an loop, but break out of it after 5 minutes.”
After you explained it what to to, the enemy already took your position and is confused what that thing in front of them is, so maybe put a bomb below it and activate it remotely…
Seems to be the least useful AI assisted killing machine possible.
It’s a tech demo. It’s not made to be practical. It is made to spur the imagination.
I feel like the next headline is going to be, “AI-Assisted Rifle Robot Shoots Hobbyist by Mistake.”
ChatGPT: “Yes… mistake…”
I was thinking about this movie just the other day. I had no idea that was Jack Black.
Is this Disney’s “the kid”
This should have the turret voice from Portal.
“Hellooo… Is there anybody there?”
“I’m different!”
“We’re under attack from the front left and front right. Respond accordingly”
slow-clap
I thought this as well… if “front” here means “Alt” then absolutely.
I’m glad I could help!
