Congratulations on the amazing work/demo! An interesting extension of this would be integrating the "Level of Gaussians" approach presented at: https://zju3dv.github.io/LoG_webpage/
Also interesting would be to generate gaussian splats from Google Street View footage, and integrate it directly into the Google Maps 3D data as you have done.
The Italians have largely voted for a party that vowed to protect their culture. Perhaps the resulting laws are shitty for you, but likely not for the majority of Italians.
In order to signal virtue, the thing in question has to be considered virtuous, by the mainstream media intelligencia and generally the opponents as well. You cannot appear virtuous if what you stand for is contrary to the mainstream opinion of how moral behavior is supposed to look like.
By deciding to give preference to history and the thing that is currently considered immoral (eating meat) they actually do the reverse of virtue signaling.
And yes, that is a pro free market move, because the only thing they say is: you can call your alternative product whatever you want and you may sell it as well, but do not try to confuse customers with historical naming that do not represent what your product is.
Vegans/vegetarians are in general for an outright ban of those products, so basically authoritarian fascist behavior; but as always, it's the most tolerant called on his behavior and somehow it is found problematic that it is forbidden to lie on products packaging...
This is a problem with many dimensions and few equilibrium points. Libertarianism and nationalism are certainly incompatible, but that doesn't mean everybody needs to make their mind an pick one. They can accept both as valid equilibria, and rally for a gradient-descent into the closest one according to local politics.
Right now we find ourselves far from any equilibrium points: just an awkward heavily polarized in-between. I hope we can find the next equilibrium without a global civil war.
PS: Not sure if "you guys" goes for me, I'm merely an observer. I'm not even Italian.
Can you elaborate or is the first word just extra weight? Please no offense, it's an honest question: I find very interesting how most people, when they want to get a debatable point across, almost always throw exactly 3 profound and all-encompassing adjectives (rarely 2, 4 or 5).
Even ChatGPT has noticed the pattern and picks up this style. But if you break the sentences down, very often realize that 1 or even 2 of such adjectives are fillers.
Same here, as if an unsolicited SMS with attachments will be effective. I also unsubscribed a while back after reading about issues similar to what's mentioned in other comments here.
They are. Our marketing campaigns are most successful via SMS because consumers haven't yet filtered out sms as potential spam yet and actually read and click. If
This already happens with Nespresso coffee machines (they have an SIM that connects to the Internet, whether you want it or not). That day is already yesterday.
Wow, they sure do. Page 29 of the user guide for the Nespresso Zenius says -
This coffee machine is equipped with M2M (Machine to Machine) techology which may be activated in due time with your agreement.
Thanks to a SIM card already integrated in the machine, such network connections will offer new services (subject to further terms and conditions) to its customers and improve the after sales process by automatically communicating machine troubleshooting / diagnostics to our Customer Relationship Centre (depending on country requirements and specificities).
What a time to be alive. This should be disclosed on the front page of the manual and not hidden in the smallprint. And it should come with instructions on how to disable it with a physical switch.
If that's the case then it cannot be used in any place/location where there is no signal.
If so, then there would be hell to pay the first time if happened. If it works sans connection, then do what I've said elsewwhere and that's to cut or short out the antenna lead.
Removing the SIM may be deemed provocative by the manufacturer, if there's no signal reception then that's a different matter (the user can't be blamed).
doubtful. I'm sure it will refuse to work if it can't talk to home base, and that home base will have some sort of certificate pinning so only their servers can authorize it do make the coffee.
> On its site, Google says certified peripherals must receive automatic firmware updates over the air and are tested for "quality, reliability, and interoperability."
What the hell. What about keeping hardware I/O thin, and handling "quality, reliability, and interoperability" at driver level?
Why do we keep adding more and more independent software stacks that require updates to our list of worries, and not less.
Sure the OS handles drivers. However, nearly all USB peripherals we use every day are 100% standardized – or at least their core functions. USB HID Class, USB Audio Class, USB Video Class.
Three things have special drivers (on Windows) in my current setup: The various USB controllers (inside the laptop and the Thunderbolt dock), the USB Ethernet controller in the Thunderbolt dock and the Bluetooth controller. That’s it. Keyboard, mouse, webcam, audio, even my wireless headset: All generic drivers.
This all means the devices themselves must implement the appropriate abstraction on top of their specific hardware.
> Keyboard, mouse, webcam, audio, even my wireless headset: All generic drivers.
Not if the vendors have their way. I plugged in an external Logitech mouse to my locked-down, corporate, non-admin permissions laptop only to see Windows automatically install a Logitech mouse utility. Not entirely sure what it does, but it also added a constantly running Logitech update service.
Given the strong security track record of hardware companies, I am sure it is fine that they get to automatically install software without any approval from me.
That’s something slightly different though. Microsoft had this crappy idea that drivers are no longer just drivers but instead you’ll also get all the usermode software you never wanted automatically.
IIRC you cannot disable this behavior without entirely disabling driver installs/updates from Windows Update.
> What about keeping hardware I/O thin, and handling "quality, reliability, and interoperability" at driver level?
Almost all operating systems come with a generic, PnP driver, that can handle most things with a meh-OK level compatibility. For example, the Windows 11 generic driver for Intel graphics causes a black bar on the bottom of my 4K screen until I install the official, everything-and-the-kitchen-sink, Intel driver for my device; but it's serviceable.
I would dream of a world where everything was handled by firmware in-device, so that we wouldn't almost ever need anything more than the basic PnP drivers that come with the OS (even though we are nowhere near that yet). You are asking why we can't go back to having customized drivers for everything. If we could have generic drivers for everything, making new operating systems for hardware would be so much easier, and we wouldn't need to fork the kernel a bajillion times for everything. Niche operating systems like FreeBSD, Haiku OS, Redox, ReactOS - they would also become so much more usable considering they'll never have the manpower to write a proper driver for every device.
On top of that... drivers rarely survive operating system refactoring. Linux doesn't have a stable driver interface at all between versions - so it's either get your drivers upstream past nitpicking and sometimes political maintainers - or keep your proprietary fork up-to-date until you get bored because you've got a new product out this year (story of almost every cheap Raspberry Pi clone board). Windows has a stable driver interface... but they still change it every few releases (Vista was a big one, 10/11 are going in on DCH), and you can't use your Windows XP drivers on Windows 11 [Note]. Firmware-on-device means that if your driver doesn't get maintained or updated, you aren't SOL.
[Note] This is also why... "what's with hospitals still using Windows XP?" Well, their software only runs on Windows XP, and often, the drivers only run on Windows XP; so even if the software could run on Windows 11, it couldn't talk to the medical device, which could cost... tens of thousands to replace.
Firmware on devices and thin, consistent USB layers is a blessing, not a curse. Who is going to port your webcam’s fancy driver to Linux or FreeBSD?
Frankly, I wish as many devices as possible worked this way. Operating Systems would be so much more secure and easier to port if we didn’t have to worry so much about those darned drivers.
Given that it is doing actual operations under the hood, it would be interesting to have it just output the "native" command for the operation rather than executing it. People could write wrappers around that output to execute it after prompting the user for confirmation or something like that so that users would have the choice to "opt into" the level of automatic execution they're comfortable with.
This way my approach when I created a similar tool, called git-genie[0]. It’s more of an educational tool first, explains the generared git command in detail.
Sure, perhaps my example was too extreme. What about:
$ gitgpt commit files with msg cleaning repo files and push
git commit -m msg
git clean -fdx
git push
Without `git-add`, 1st one is a NO-OP, 2nd is a destructive action, 3rd is a NO-OP. All vaguely related to the topic at hand, "hallucinating" such a destructive action seems at least plausible.
The GitHub Copilot CLI tool that is in beta, along with other tools, will show you the command and then make you manually choose to run, redo or abort.
In practice I have yet to accidentally perform any destructive operations. This makes intuitive sense because those seem like unlikely completions for a model that has been fine-tuned on “helpful cli recommendations”!
Extraordinary claims require extraordinary evidence. Given that GPT3 has 175 billion nodes, how would you even begin to support the claim that it never (or sufficiently rarely) does things that are surprising to humans?
But you can get a general feel by using ChatGPT. Open up a new conversation and ask it something like, "What is the capital of France?". Note the response. Open up a new conversation and note the response. Soon enough you should be able to see that the responses are far from random.
You can use the OpenAI APIs directly and have it run 10,000 or so iterations to see what kind of "hallucinations" it makes! They are not random!
Ask it details about a little-documented event and it'll happily tell you plausible, but utterly false, lies, however.
Apparently the "early 2011 Bougainville earthquake" was magnitude 6.3, at a depth of 21.7km, on the 20th January and caused "widespread damage to buildings and infrastructure in the region, and triggered landslides that blocked roads and hampered rescue efforts".
It was actually on the 7th Feb, a 6.4 and at a depth of 415km. There were "no immediate reports of damage or injuries".
None of this is remotely surprising, considering it's a turbocharged statistical model and it probably ingested a few words about it at most, out of billions and billions, but somewhere along the line from "famous" to "footnote" subjects, it will segue into complete fiction.
Some flavour of "git gc" after your reset is far more likely to crop up and ruin your day, that's true.
As long as you stay on the statistical beaten path (i.e. you're asking about Paris), you will probably be fine, indeed. Probably. Stochastic bugs are always the most fun anyway.
You definitely go about making “stochastic bugs” more reliable in a manner different from debugging software.
It’s more akin to industrial engineering. There is no such thing as a perfectly machined widget. So we come up with an acceptable range of tolerances and compute a process capability. Six sigma. 3.4 defects per million and then buy an insurance policy.
Thanks for the link! I don't think that really addresses my concern, though.
My point is that these LLMs are basically incredibly large programs that defy analysis with our current tools. Sure, I can poke it a few times and see that it usually does what I want, but that's not the same as saying it never goes off the rails.
If it does something crazy like post my bank login online, even only once in a billion times, that's still orders of magnitude higher than I'm willing to accept.
You’re basically asking me to prove to you that I can’t fly.
I will say it like this: it is highly improbable that I can fly. I cannot come up with a way to prove it to you. There is some sort of epistemic miscalculation going on if you operate under the assumption that I might be able to fly.
I don’t disagree with them. Reality disagrees with them.
When LLMs “hallucinate” they don’t do so in random ways like somehow deciding to “rm -rf /“. They do so in predictable ways.
What, am I supposed to handhold every person in these forums until they stop reading garbage on the internet and start either doing their own empirical research or at least read other empirical research?
Edit: Apparently the answer is yes, I am supposed to handhold every person in these forums… and I will tirelessly do so, even in the face of provocatively incorrect yet populist statements like OPs… I still imagine I lose my patience every 100 or so instances, as seems about the average these days…
If you get tired of respectfully explaining why people are wrong, then you can always stop commenting and go get some fresh air. What you should not do is lash out aggressively with zero explanation for why you feel so strongly about this.
I didn't intend to patronize, just remind you to chill out. I need the same reminders from time to time.
I note that the conversation that started from your re-submitted (and much improved) comment was quite productive and insightful on all sides, which is exactly what the site guidelines are designed to enable.
Wrong, perhaps, but not aggressively so. They were making an observation that could very well have been misguided, but there was no aggression in their tone.
Look at the next response where OP admits they didn’t need to be so extreme.
This is of course to be expected based on those that are optimizing to be rewarded primarily for their opinion and not for how it is stated.
I don’t find your argument convincing because I am not purposefully ignoring the social context. Your intentions are suspect, even if they are subconscious.
Like, if I’m unreasonably polite in response, downvotes. If I’m reasonably impatient, downvotes.
Also interesting would be to generate gaussian splats from Google Street View footage, and integrate it directly into the Google Maps 3D data as you have done.