fennecfoxy

Never seen Taalas?

Let alone a decent sized model on a couple 4090s or similar, pretty gud t/s to churn out tts/stt/response/control model actions and emotes.

Nah weaboos and (us) furries been doing it for ages + not necessarily expensive at all, can use a webcam for basic ones for ages now.

https://imgflip.com/i/ax3s0h

Lmao, really? England has famously been judged for its industrial colonisation, even if it was still in an era where _everyone_ was colonising _everyone_.

Interesting how Genghis Khan got away with it, to most he's now just a "badass" historical figure, I don't think most people could tell of all the terrible things he was responsible for.

Nope, but damn bro chill tf out.

This is definitely something I've always thought about doing once I retire (if my generation gets to retire).

Just seems like the ideal way to spend the last period of your life; quietly making the small mechanical pieces and hopefully finally assembling something to be left behind.

Though I don't imagine I'd ever be able to produce something as small, accurate or intricate as these students are able to.

Why?

Many modern watch parts are CNC machined, often the finishing is done by hand such as zaratsu polishing - but even that is a repetitive motion that can be mechanised.

I would not be surprised if given enough time even what we have today - a decent VLA model + some other specialised models, 6-axis CNC machine, an SMD pick and place etc would be capable of designing, manufacturing and assembling a mechanical watch.

Is it? Android tap and hold/image text select one-shotted it in 2 seconds.

But it's not hard to OCR? And I don't know why the article dedicates an entire section to it.

On a Samsung S24U I held down the "circle to search" homescreen button which brings up the AI tools interface (I don't know what it's officially called), held down on the text and copied the whole thing in one shot.

It took like 2 seconds.

That doesn't seem like a reliable method at all.

Not really odd. Sure you can get the same with a Google search but all of the attention is on AI atm so you don't expect it to be unfairly judged?

Humans aren't like that lmao. We're reactionary, tribal animals.

So damned if you do, damned if you don't, right?

I can totally get the "don't tell people to kill themselves" aspect of these models, but certain parts of the Internet have been telling people that for decades.

Certainly the models should be trained/tuned to avoid conversations like that wherever possible and redirect people to get the help they need...but that's exactly the problem; doing that is MORE than what the state and the strangers surrounding a person would do. That's the problem, a mental health crisis that is ignored, particularly in men.

Not to mention missing out on new features like Roborock's rover, etc. That's not gonna come to anything DIY (well DIY from a kit...is that _really_ DIY) any time soon.

And also quality.

I wouldn't think that the copy of some movie Netflix is streaming to me will be 60-100GB over the duration of the movie. Not to mention when their services have issues and you're watching 5-10 minutes of low quality content until it settles and snaps up to full (streaming) quality.

Millennials never spent like >6-8 hours a day on phones doing nothing productive.

Yes, as adults we now spend loads of time in our places of work on devices but generally in the pursuits of doing productive things.

And while I haven't looked up the stats, I don't feel like it's exceedingly more dangerous for kids to play outside today than it was back in the 00's, 90's, 90's etc. I just feel as though we're more aware of the dangers of what _could_ happen to kids - all of which is a facet of the media/social media pressure and scaremongering.

Plus there are still loads of feral kids allowed to roam around and be little shits anyway.

I mean I believe in protecting your company's IP, but IP and patent law is absurd these days, designed to protect investors and their fake money rather than actual inventors (who usually get no proceeds/are shafted).

They trained from the internet, so if someone trains from them it's fair game. Their clever tech should be in the mechanism with which it uses to provide an answer, not the answer itself.

Why? You're being sexist and I hope you can understand why.

Yes, for actresses _and_ actors I'm sure you'd get the same level of performance as you would for any facial recognition use case. You can't do facial recognition on someone's back, but I'm sure there are other techniques/models that can be applied, many people have unique marks/features etc.

Ha ha ha, it's because most humans overlap on a few things - like eating, shitting, sleeping and fucking, ha ha ha.

Seconding runpod.

They were having availability issues with GPUs (of course) but especially their UI where you'd customise a template only to try to start a pod, the GPU be unavailable and the UI reset forcing you to make the changes all over again.

But they have fixed that since, now starting a pod is more from a live page where as GPU availability status changes it updates in realtime/if your deploy fails you just try again - your customised env vars etc are still there.

Plus they also addressed the GPU availability problem as something they're working to fix and it's understandable seeing as nobody can get their hands on GPUs atm.

Look at how many fuck ups there are with basically no repercussions; the dude is still rich.

Is any more proof needed that if you ride the coat tails of real engineers and have a propensity for screwing other people over that that is what success really is for a lot of people in business?

Just look to the number of huge companies with founding members or integral early members being forced out just as things are getting good ($$$).

"Oh my God"

Lmao I bet Dang is watching this chain like *finger on edit button*

If only because we're too apathetic to make laws and enforce punishments for the rich and powerful.

People have always been able to be spied on.

But then I watch people speed all the time and it seems to be a common human behaviour to break the rules to the detriment of others when they think they can get away with it. It's disgusting.

Yep, web serial okay but not bluetooth? "We can't make it secure". Yes you can.

No web bluetooth from inside an iframe, always require consent to select a device and connect to it.

Sure, I suppose it could enable more scams by tricking people into connecting to x device but if we were truly worried about stopping scams we'd turn the internet off and then shotgun our entire species.

Probably quite a lot - if you look at what Anthropic found around persona vectors; https://www.anthropic.com/research/persona-vectors.

I imagine the context will always sway the model to some degree, not only for the task you're trying to get it to do (aka instructions) but also its persona, how accurate it is and the way it acts.

Because governments and councils waste money on paying private contractors do do stupid bullshit like installing statues or gardens or other vanity crap meanwhile grass areas are overgrown and the roads are filled with potholes.

The general populace aka the voters are too apathetic and absorbed in their little consumer lives and tribally motivated political quibbles to know or care that so much tax money is wasted.

I quick Google search reveals for my Samsung S24 Ultra: settings->security & privacy->Permissions used in last 24 hours->location->historical log: Google (background), Google Home (while using app), Maps (background), etc.

Query: context: { "boss_email": "bigboss69420@corporatepersonhood.net", "upcoming_meetings": [{ with: "bigboss69420@corporatepersonhood.net", "time": "11:00" }] } user: i need to contact my boss i will be late, could you tell him I'll be 15 minutes late?

Output: [{"name":"send_email","arguments":{"to":"bigboss69420@corporatepersonhood.net","subject":"upcoming_meetings","body":"I'll be 15 minutes late"}},{"name":"send_email","arguments":{"to":"bigboss69420@corporatepersonhood.net","subject":"time","body":"I'll be 15 minutes late"}},{"name":"send_email","arguments":{"to":"bigboss69420@corporatepersonhood.net","subject":"time","body":"I'll be 15 minutes late"}}]

Context definitely helps. But yeah the quality of it doesn't seem to be too high. To be fair it makes you realise that not only is parameter extraction required, but also content generation (email body). Also debouncing the 3 tool calls.

Maybe under very specific circumstances/very tight harness this sort of model would be useful?

Context is everything