This doesn't seem like the right way to do business long terms. The off chance that someone actually take you up on it and pay your 'bill', you've destroyed a lot of goodwill and alerted the rest of the tech world of your scammy moves.
Also for those that require a credit card for a free trial, I always use a virtual card and cancel it. It's super fun to watch them cry when they can't actually charge you.
Funny, I got a fraud call recently because CrunchyRoll decided to try to renew a subscription I abandoned years ago and the card they have is expired.
I know it wasn't me because I gave up entirely on the service after they changed something about their login systems to reject my password and I could no longer get in. Support wanted me to jump through a lot of hoops and I just refused, choosing instead to just stop doing business there because I wasn't really watching anything at that point anyway.
This was around 2022, mind you, so they tried to renew me after several years with no explanation.
It's way easier to just not give them a way they can charge you. That way you don't have to deal with a support representative fakely asking you how your weekend was, and who doesn't actually care about your weekend.
I'm surprised this is even a thing. After all, you go to Google not for the truth, but to search Google. Since when is truthiness the "guarantee of service"?
You're not even paying for a google service, search is free... You might be the product, and your data, but you didn't directly pay for a service and they didn't sell you a fake service.
I'm not taking Google's side, this isn't about whether it's right or wrong to rob websites of traffic, this is about AI's returning search metadata.
But I'm surprised that they lost this argument, and the line they took in the first place.
The Internet isn't made of fact checked data, it's crowd sourced. How can anyone be liable?
That is exactly the point of the ruling, though... they are saying that AI summaries are NOT the same as search. If Google was just returning search results, and then users clicked on a website and read the content there, Google is not responsible for the content.
If instead Google gives you an answer right there on google.com, without going to another site, they ARE responsible for it.
Not precisely. The issue at hand isn't just that Google displayed the AI summary, but that they authored it, making them responsible for its contents. If the defamatory content had been in a snippet in the search results, they would've been fine, because that clearly has another author who can be held responsible. The AI summary has no other author than Google; therefore, they're responsible for what it says.
(What's the alternative, after all? Having no one responsible for what the AI summary says is clearly untenable.)
What don’t you understand? Those websites that defame a company are liable for that defamation. In this case Google defamed a company in its AI summary and is this liable for that defamation.
but if Wikipedia itself writes harmful content such as encouraging people to drink bleach, then wikipedia is liable. Google now generates its own content with AI, that defame others, so Google is liable.
And those tons of websites are liable for their misinformation. It's probably not worth suing some random blog because the author probably doesn't have money or lives in Russia. But Google has lots of money and a legal presence in almost every jurisdiction.
That's the difference between returning search results and interpreting the information and summarising them. If a newspaper says 'so-and-so has been arrested for theft' it's not the same as them summarising to 'so-and-so is a thief', the second is potentially libel. Why should Google be held to a different standard?
Google itself is more trustworthy from a normal person perspective as they use it a lot.
None of "AI" companies call their apps "Entertainment fun text generator". They are call them serious names, use words like "intrllegence" and "thinking".
So yeah I'd think if any of "AIs" start to recommend to drink some bleach or take a flight from a 10th floor window these companies should be liable.
I think it's very clear that Google's AI overviews go far beyond just searching Google because they often incorrectly compile sources to come up with an incorrect answer. For example of this look at the comment I made in this thread
I go to Google to search, but get spammed as if I wanted to talk to a chatbot (and a very poor quality chatbot at that).
This is a gigantic own goal for Google. The average person’s impression is that Google AI is much worse than ChatGPT, even though that’s not actually the case. Google is shoving a terrible model in everyone’s faces.
The question is whether Google is publishing false claims or relaying other people’s false claims. The court found it to be the former which makes sense to me.
However there is justice for commenting poorly, there doesn't seem to be justice for posting lies and deceit, which is borderline a serial case here(?).
That question does comes up quite a bit, so I try to occasionally post an in-depth explanation. Maybe I'll do that here.
We can answer the question by rephrasing it. I would put it this way: "is it ok to be wrong on HN"? The key change is to replace charged words like "lies" and "deceit" with the neutral word "wrong".
People do lie and deceive, but to tell that apart from just plain being-wrong requires knowing someone's intent, and this is not something that moderators (or readers in general) can do. Even courts of law have a hard time doing that! HN comments don't contain nearly enough information to decide it. You would have to have a mind reader [1].
Worse, readers are far too quick to assume lies, deceit, etc., in other commenters that they happen to dislike or disagree with. The internet is rife with this, as we all know. Most of these insinuations are imaginary—we are all too inclined to attribute disingenuousness to the other side. One can easily see this by looking at how the same accusations are made against one's own side.
Once the question has been rephrased in this way, it answers itself: of course it's ok to be wrong on HN. How could it not be? Most of us are mostly wrong about most things. Moreover, being wrong is part of eventually getting something right. Curious conversation, which HN exists for, often involves mistakes, and sometimes the mistakes (or seeming mistakes) turn out to be creative.
Even if we decided it was not ok to be wrong, how would we enforce that? We don't have a truth meter [2].
As we have neither a mind reader nor a truth meter, we can't base moderation on anything that requires knowing people's intent or knowing the truth about things. We need criteria that can be decided from people's observable behavior on the site. As I sometimes put it, we can only moderate by effects [3].
That's why HN's guidelines are they way they are (https://news.ycombinator.com/newsguidelines.html) - there's nothing in there which requires us to know what someone's intent is, nor whether they are right or wrong on a topic. At least I hope there isn't!
What about the truth, then - do we care about it at all? Of course we do. It's critical! But working out what's true vs. what's false is the province of the community, not the mods. It's the commenters' job to do that, e.g. by answering bad arguments with better ones and false information with true. And they should do this within the site guidelines, e.g. by being respectful and curious rather than aggressive and accusatory.
The moderators' job is to hold the container for this or (if you like) to keep the playing field fair. This isn't possible in any complete way - there are far too many factors pushing things into messy, unsatisfying places. But we do what we can.
Well actually, spacex owned xai is now coding AI in C while earlier it was rust; I guess with the attitude that a language good enough to control rockets is fine for ai.
At this stage just expect that every accounts will get leaked or rooted, it's a matter of when, not if...
Use varying email `plus addressing` (john+am2604@foo.com), varying passwords or passkey and 2FA on anything remotely important (use of your identity, not just financials).
Plus addressing (or movable periods in gmail addresses, etc) is increasingly pointless for a whole host of reasons.
It may keep out the bottom x% of spammers/hackers but it doesn't do much for the increasingly sophisticated scams that are appearing.
If the bit before the + ends up in your inbox anyway then it'll just get stripped off and used. Spammers seeing this kind of thing across several breach dumps:
and will leverage that to target spam at you for other sites, or just email bob@example.com as there's a good chance that'll get through.
Years ago I did a test with my own domain where I created who unique aliases with plus addresses, e.g. steve.smith+iawer@example.com, bob.jones+wpoqe@example.com
It didn't take long for emails to start arriving to steve.smith@example.com and bob.jones@example.com even though that email address had never been used anywhere ever before.
As others have said, you're better off just creating unique emails with `pwgen -s 16` such as wmR5pNhGI8yidU7N@example.com and storing that in your password manager alongside a similarly random password. (Yes, this is roughly what those unique email address services provide.)
Also many services/sites/providers simply assume the username is immutable. $DEITY forbid you might have to change your email address at some point in the future.
I recommend people use proper email aliasing, not plus addressing. Duckduckgo makes a free one that's can integrate into Bitwarden, if you have iCloud+ Apple's($0.99/month) hide my email is good. Addy.io and SimpleLogin are the best and allow PGP encryption to prevent another party having access to your emails, but they are paid for full features.
> Organizations like the IAB require that advertisers normalize email addresses so that they can be correlated and tracked, regardless of users' privacy wishes.
I sympathise with the author being in the same boat, largely.
I just want to emphasise a point... Calculators give 100% correct answers and yet we still hire accountants; for the simple fact that we don't want all to be accountants.
People will hire software engineers for the simple fact that they do not want to be software engineers.
Computers came in and "took" the job of calculating numbers (I assume usually budgets and finances), but instead of every layman just using a computer to organize their company's finances, they still hire a professional to use the computer to organize the company's finances. The role shifted, but it wasn't eliminated.
there are literally no more people "computing" by hand. job as defined by "perform calculations by hand" is totally gone. none. nada.
this "shifting-role" rhetoric is very dangerous. making definitions fluid is a very slippery slope. you can arrive at any conclusion you want and support any point you want by changing defintions. seeing it in AI from C-level leaders is very concerning.
With todays LLMs, yes. But if they can ever reach a level of a contractor in a reliable way and companies offering them willing to take responsibility (because confidence is high and rest is insured), then one can just hire a cheap AI agent to fulfill a contract - design, implement, deploy, run and maintain your service/website like the engineer before.
Calculators are not a replacement for accountants, online accounting services are in many cases. Which again can be run by an AI if they reach that level of reliability.
People don't make their own bread. They buy it from an expert.
But bread shops are available on every corner. Will software jobs become as common as bread shops? If yes, what happens to the salaries? Something to think about.
Of course they can, but if you give a LLM a specific rule, what happens is that it only shifts the probability of following the rule. Prohibiting a rule violation is technically impossible via prompting.
Humans do make mistakes or forget things as well. We learn to not rush on stairs and to not touch hotplates. A few bruises later that wisdom is permanent, at some point we don't even need to fail with everything to accept the rules.
A LLM is permanently at risk to break every given rule.
I think humans learn this from much more than documentation.
We learn it from conversations with our managers and peers. We learn it from reading between the lines of those conversations. We learn it by being in meetings and seeing who is reliable and who isn’t. Etc.
LLMs do great learning from documentation, but so much of what it takes for an employee to be successful isn’t and can’t be documented
In the US, all you need to work in tax prep is a high school diploma and most individuals are not worth the cost of an audit.
I wouldn't say it's particularly brave, in fact LLMs are probably better at identifying mistakes than most tax payers. The % of Americans using a CPA to file taxes is fairly small.
you know i had a tax accountant that said i dont have to cap gains on iso exercise . i checked it with ai that gave me all the detailed info that i took back to accountant who then agreed that they made a mistake.
Even if you dont feel comfortable using ai for taxes. you can atleast understand your situation better. And go to accountant for final 10percent. It is not really that hard.
That's fine if you have a relatively straightforward situation. If the situation is a bit more complex, with important decisions you should consider, it might still need some eyeballing and time. That is accounting.
Herb's blog post links to the SlashData Developer Nation Survey, so presumably that's what the claim is based on. The company has a methodology page here [1], and it looks like the Developer Nation panel [2] is one of the sources used by that company.
Hi, I'm a Principal Market Research Consultant with SlashData. We've been around for more than 15 years, and we work with the largest organisations in tech when it comes our market research on developers (our client page include AWS, Google, Microsoft, LF, Cisco, and so on and so on). Not bragging, but trying to assuage concerns that we are some fly-by-night org that popped up out of nowhere.
The makers of the documentary reached out to us to use our footage in the report and Bjarne has also used our report/blog when talking about the future of C++. In general, our hypothesis about its continued relevance is in line with what the documentary itself reports: developer numbers have grown massively and C++ has a very specific series of use cases that maintain its relevance.
Our methodology page is lengthy, but it can be summarised by the combination of several pieces of distinct information. We have labour statistics from various national governments to provide reference points, we use statistics from areas developers associate with (e.g. StackOverflow or GitHub) and compare against responses from inside the survey, and our own build-up of numbers based on the proportional data from within the survey (region, age, area, development, etc.). We have been doing this for many years. We also acknowledge in our reports that we are confident in all regions, but with greater uncertainty in the Greater China region, as we are a non-Chinese firm, so we have to work with partners in the region which minimises our confidence in the region.
The report lists the numbers for the major leading languages, basically any language that is used in multiple areas of development and has more than 500k users, give or take. It is not a perfect measure approach but it is about giving the most useful information at the highest level. We also exclude things like SQL because they are nor programming languages, unlike most of the other measures.
For the report, we typically have blogs available with similar information, but our free reports are designed for lead generation, which is why emails are asked for.
It will be interesting to see the first AAA game that uses these methods instead of rendering a 3D world. Even if made from CGI worlds, it would be a very interesting approach and with somewhat predictable performances.
Reminds me of Ecstatica [1], a 1994 game that had intense visuals with a very odd/different rendering engine made of 3D ellipsoids; in a way really crude splats in gouraud shading.
I know this comes up a lot on HN because its not primarily a graphics community but:
1. Gaussian Splats are very expensive to render. They capture a lot of detail which makes them seem cheaper than an equivalent raster render of that quality, but they wouldn't meet real time AAA game performance requirements
2. Gaussian Splats don't have a concrete surface. Want to cast shadows or do physics? It's doable but very tricky. Want to relight them? Also tricky. What is the exact surface point that you want to affect or sample for any particular operation? Deformations also become very difficult to do well.
3. Gaussian Splats are not sharp. You can get sharper with different kernel types or higher density of points, but your costs go up as well.
4. Gaussian splats are awful for any kind of path tracing. You can do it but you go back to the issues above. So mixing and matching traditional content with splats becomes a performance bottleneck.
I don't think you'll see a AAA game use splats for more than something like cinematics in the near term.
I'm working on the vision component of a drone racing stack. Could I use GS to render my living room as a digital playground to train my vision models in?
I know nothing about the technology but the alternative is creating a 3d model of my living room which is also outside my skill-set.
> Could I use GS to render my living room as a digital playground to train my vision models in?
Yes, its what the autonomous car people are doing.
However you might want to do photogrammetry first (https://github.com/alicevision/meshroom opensource) as that produces a mesh which you can use to detect collisions easier. The downside is that transparent objects render really badly. but it is a lot faster to render
Ehhh if it's just for looking and you don't have anything lidar just go for splats they're way better behaved, mostly because they don't need to understand a concept of "surface" they just understand "splat with spherical harmonics of view-dependant color".
Splats would work and it’s what a lot of automation folks use. But the risk is that your splats aren’t tight to the surface and that can cause false positives.
For training you can do a hybrid geometry plus splats workflow. Have geometry that you can constantly raycast against and have as an input to your vision training or to get accurate depth buffers.
The workflow for splats and photogrammetry are very similar.
The contributions of 3DGS lie in how fast you can make them in modern GPU hardware (tiling + sorting with threads), and how to make the pipeline differentiable so that you can fit the Gaussian splats with photogrammetry data. Similar to the history of deep learning, it became technically feasible once the GPU hardware was powerful enough.
Yes but in a weird way (at least on vanilla 3DGS) - it duplicates the splats on the mirror side.
Vanilla 3DGS cannot do any specular lighting or reflections - the color is basically baked in the splats. There's some active research going on to create richer Gaussian splats so we can do shading (or even ray tracing on it) - but haven't seen anyone using in production yet.
Many years ago there was a game called Casebook[1], a small little detective game where you investigated rooms for clues. But unlike similar FMV games where you jumped from point to point, it had photorealistic environments that could be smoothly walk around in, much like later lightfield or gaussian splatting experiments.
I can't say I know how they actually did it, but taking a look at the trailer I can point out that it looks like the spaces are confined and your character is on rails. I'm mainly going off of the instant direction changes that don't appear to be 45 degrees off from the camera direction. Once it's constrained down to a single line/path you could do some wild things like cube mapping a video, where the position in the video is tied to the characters position. I can't say I know how they would take that video though, my best guess there is the scenes are constructed in 3d software, just it was to expensive for real time rendering.
Cube mapping a video sounds plausible, this is commonly known as 360° video. Putting the camera on rails (though I don't really notice rails in this case) and tying the video playback speed to the speed of the rail movement has also been done in the past in some pre-rendered PlayStation games, though without cube mapping. But I think it's not pre-rendered in this case. It looks far too realistic for a game that is at least 17 years old. My best guess: they captured the 360 degree videos with a real camera (stabilized in some way) and edited the equipment out frame by frame.
The image capture was done with a robotic camera rig from what I understand, they photographed 360° images of the room from all possible position. They restricted the camera movement to a plane, which is why the player height is fixed. I don't know what they did on the software side with all the image.
Oh cool, so the camera wasn't just on 1D "rails", it was on a 2D plane. I never before heard of a game (pre-rendered or photographed) which did that. Impressive.
Dreams for PS4 used point splatting and has a very unique look as a result. The splats were created from distance fields instead of being scanned, so they don't look like modern gaussian splats. They have a painterly look instead. https://youtu.be/2ltgkcoQzow
This is "rendering a 3D world". It's basically the exact same techniques that traditional rendering uses, just with a different primitive that's not triangles. Everything else pretty much carries over.
If you mean the technique of splatting specifically, Dreams for PS4 [1] is prior art.
If you mean pre-rendering, there's Myst and games like the original FF7 for PS1.
It's not gaussian splatting, but Outcast (1999) has an interesting voxel-like rendering for the world surface. It has a pretty distinct feeling when walking around in the early areas, and a somewhat clunky but usable UI.
> The game does not actually model three-dimensional volumes of voxels. Instead, it models the ground as a surface, which may be seen as being made up of voxels. The ground is decorated with objects that are modeled using texture-mapped polygons. When Outcast was developed, the term "voxel engine", when applied to video games, commonly referred to a ray casting engine (for example the Voxel Space engine). On the engine technology page of the game's website, the landscape engine is also referred to as the "Voxels engine". The engine is purely software based; it does not rely on hardware-acceleration via a 3D graphics card.
Bladerunner: Revelations used a similar technique to bake down large CGI worlds with expensive lighting into something that ran on a Pixel 1 at VR specs.
Its honestly really very hard to work with this stuff because you ultimately need to be able to meshes inside these scenes triangle seas and you need to do it in a way that plausibly fits in the world. You can't have unlit characters walking around a baked lit scene and have them fit in. That's just from a visual design perspective.
You also always want to have bounce light from your dynamic things onto the baked scene and depending on the tech, you might not even be able to spatially place a dynamic thing and have it properly occlude what splats it needs to occlude.
As is, its a niche technology for games. That might change one day.
I think it's inevitable it goes there. Right now the level of detail and quality of games is limited by the console/PC hardware you're playing on. But with the splats they can render the whole game's world in a massive server farm at Hollywood Movie quality. I imagine there might be some balance of splat and traditional rendering technology since not all objects will lend themselves well, but this might be truly transformative.
There wouldn't be any cloud. Splats are still local, but all the lighting and texture are pre-rendered. The problem is they're not interactive, so they'd be good for a lot of the environment but your main character and other things that need to be interactive would need to use a different approach.
reply