Posts by DrXym 5627 publicly visible posts • joined 18 Jul 2007
Samsung has been making folding phones in a variety of form factors for years now. And as bad as I think folding phones are, I think it rather more likely Apple copied Samsung. Also to be fair, the Pixel Fold and Oppo had been playing with more squat form factors before that.
But all this is beside the point. The reality is Samsung was there first and then a virtually identical phone shows up from Apple.
When Samsung first launched the Samsung Galaxy phone, Apple were throwing sueballs around because it aped their own iPhone design. And now Apple are copying Samsung. This "passport" folding phone is basically just a Z Fold 8 with rounder corners.
I'm sure fools with more money than sense love these folding phones no matter who makes them. Who cares if they're basically e-waste - got to have the latest and greatest thing.
Legislation has to assume and mitigate for the absolute worst manufacturers behaviour and incentivize them to comply.
Because of course they're going to be dicks about it otherwise. Phone manufacturers want to sell new phones and turning their old phones into e-waste is an easy way to do it. Even if they were complying with the rules about repairability they could limit supply, or use DRM, or create onerous part pairing requirements, or use proprietary connectors or any other tricks they can think of. We've even seen how they've lobbied to carve out exemptions if batteries perform for a certain duration of time, and the phone is water resistant.
So the EU needs to shape the rules correctly and come down hard on offenders. Otherwise the same thing will happen over and over.
Even if there no AI in libreoffice there could be an MCP like interface that allows people to plug a LLM of their choice into the tools to do stuff like document generation, grammar / rewording, citations, synopses, translations etc. That LLM doesn't even need to be in the cloud, it could be something like LM Studio.
There is no "guarantee", but you could perform defence in depth. Generate SBOMs and check for vulns. Use a proxy for fetching deps since some of those also scan for vulns and policies can be set up to distrust packages based on rules. Also use audit & socket scan commands in your cargo / npm / pip periodically and also from CICD. If you use Renovate then set that up likewise.
On top of that you could security harden the build system, e.g. containerize tools so they can access the proxy but "phone home" code in postinstall / build.rs shenanigans could be prevented.
And these days you could also throw AI at the package dir from time to time to see if it can detect malicious code.
More likely the code is just shit and either leaked memory until there was none left, or blew up because of some situation it didn't handle properly. It would be interesting to know if it could be fed crafted data from a malicious RFID that lead to some kind of exploit like bypassing checks or to capture biometric data from travelers behind who use the same kiosk.
So how many people shop in Sainsburys on an normal day and how many innocent people are falsely flagged as a result? Because for a supermarket this size there must be dozens of false positives and "trained" managers accusing confused customers of being shoplifters. If they want to use these systems there should be a counterbalance, e.g. the entitlement of anybody falsely accused/confronted/ejected by the system or staff to take their complaint to independent arbitration with £10,000 cash compensation if they are found to be a victim of the system.
Please put your name down for one of these trips.The potential for catastrophic failure or a slow suffocating death is probably not that great. And look at Titan: if it does go wrong, people won't remember you as a rich asshole whose death is a monument to hubris. They'll mourn you deeply.
Okay so it's nice you can take this model and reuse it. But you have no idea how it was trained, what deficiencies or biases it has, what booby traps or deliberately misleading information it has, what guardrails or other aversions to answering questions it has. All you know is "here is a blob of binary" that seems to work but you do not know how or why.
To be clear by garbage I mean plausible misinfo, in the style of that proposed using fonts. i.e. coherent, readable sentences that almost make sense but mislead the AI into ingesting it, poisoning it in the process. Of it's consuming images then throw some fakes in there too. Depending on the nature of the website and its general trustworthiness this could do a heap of damage to the model.
Maybe the better way is to have a surreptitious "authentic user" test, e.g. detecting for mouse movements or humanish navigation, and if it isn't received, then insert garbage into the page. It could even be styled so it's not visible usually but be garbage all the same. Maybe even use AI to generate the garbage.
If I were in the business of destroying Flock cameras I would be considering other means than this. The courts & cops are far more motivated to pursue arsonists and people shooting firearms than they would if somebody fired a water balloon with resin at them. The net result would be the same, and yes it's vandalism/property damage but the potential criminal charges are far less.
The real pros might extract one from its pole for analysis and figure out where its weaknesses lie in communications, software or hardware and do a class break rather than attacking them one by one.
I'm sure you can "distill" knowledge from one AI to another if you know what to ask but I bet the distillation feels a bit blurry and vague like photocopy. It conveys the same information but it has captured errors, as if creases, paper clips or your finger were captured in the copy.
I doubt Airbus is any different from anywhere else - someone will ask why spend a large amount of money to rewrite something that already works and that will be the end of the discussion. It will only be if the CTO orders every department to identify all of their processes that involve spreadsheets and then goes through each of those with a critical eye. Even then most of them would probably be left the way they are.
I'd agree but at the same time I could see some sales/engineering department of Airbus having some stupidly complicated spreadsheet that takes a raft of inputs to visualize something and the thing has grown organically over many years sprouting arms and legs, and nobody has the time or desire to replace it.
Assuming they do it does not hurt to try the alternatives and see if they function as expected or not. And if they don't, then they could put a bug bounty on it.
AFAIK, Excel and (LibreOffice) Calc have virtually identical maximum limits in terms of rows, cols, sheets. Chances are that either would die a death before you ever got close to testing those limits. The biggest danger of taking a spreadsheet over are things like VBA, undefined behaviour, power queries/pivots, rendering of graphs and some missing functions.
Same here. I would say my takedown success rate is perhaps 30%. YouTube doesn't care and should be made to care by becoming financially and legally liable for the scams they stream for money. If I can spot a scam just by looking at a thumbnail or a few seconds of video then they sure as hell be able to even if they have to go to the bother of training AIs to spot signs of scamminess.
That kind of scam is always maverick engineers who disrupted an entire industry with their revolutionary gizmo. Where said gizmo is just some tat shat out by some Chinese factory selling at a 4x markup over the same thing on Aliexpress. Air cons, heaters, one million lumen torches, security cameras etc.
YouTube and other social media platforms run scam ads with impunity and make it difficult and usually futile trying to report them. If they were suddenly on the receiving end of massive fines for neglecting to monitor and vet ads they might actually do a little due diligence.
I've used Claude Code a lot and my experience is like a coin toss - sometimes it gets things bang on, or near-enough, other times it totally misses the point or does something superficially okay but adds technical debt. . You really need to know what YOU are doing to know when it doesn't know what IT is doing.
I do find that it works better with an existing codebase where it has context and rules to follow. AI is also useful for donkey work - I told it the other day to write some line intersection tests against a complex shape with a bunch of scenarios and it just did it and saved me an hour's work. I've also asked it why a bug was happening and it found the fairly obscure reason with some coordinate code where I was using a y instead of an x.
So AI does have uses and it does augment understanding providing you pause to understand. But letting it loose in a massive project with minimal guidance is asking for trouble. Even if it passes all tests, it will be a unmaintainable mess of code which cost a fortune to wrangle with AI.
If it was 500,000 lines of code originally you're looking at probably a 6-12 month project at least to convert it by hand. Probably for 2 or 3 people including test cases. But there is no doubt in my mind that AI could have done a better job if it were properly supervised and I wouldn't have much confidence that somebody telling an AI to shit out a port in 11 days is supervising much of anything.
Perhaps it is broke but not in the way you think.
There are a lot of tools written in Javascript and every time you run them, NodeJS has to kick off, parse the source, execute it and suffer all the issues of a high level language. That's a lot of wasted time, memory overhead which translates into cost.
So some tools are being replaced with natively compiled equivalents. For example Babel is being replaced with SWC (Swift Web Compiler) which is written in Rust. And even the Typescript compiler is getting a rewrite in 7.0, in Golang.
So if Bun is some honking big tool written in Javascript and every time you invoke it, NodeJS kicks off and parses a bunch of source and puts an overhead wherever it runs then it becomes a problem. That does not mean allowing an AI to shit out a compiled version is necessarily a good idea either. From the way it was generated it sounds like a ludicrously rushed attempt. A more curated and guided port with AI might have taken a bit longer but lead to code that humans and computers alike could maintain.
This effort has been going on for years so I wouldn't say they're rushing anything. Yes they've found bugs and they're striving for comprehensive test suites to ensure compatibility. In the process they've actually described undocumented behaviour and bugs in GNU coreutils. The tests benefit both projects since they have a basis for comparison.
As for the license, I don't see it being a big deal on Linux, which already copious amounts of MIT licensed stuff. But it may benefit other platforms - Windows, BSD, Android etc that they gain comprehensive compatible equivalents of coreutils.
Why? Because instead it allows software to hint where it wants to go, and any special groupings and leaves it up to the compositor as to how to apply those hints. The compositor can also remember the last place the window was and restore to that place without the client asking for it it or having special knowing. That means kwin remembers where windows were and where windows would like to be and allows for more consistent behaviour on multi monitor displays where somebody might be plugging or unplugging a machine from a dock, rotating screens or whatever.
i.e. it's a feature, not a bug.
They hope they can focus on the cash cows and lose the rest, or force them to become cash cows too. The problem is this only works short term. By being so awful they're forcing companies to make a decision and clearly some of the largest, most lucrative customers are choosing to walk away and sue them on the way out. And the ones that remain will have their eye on the exit from this point forward. So short term gain, long term loss & reputational suicide.
On the one hand, being dicks might scare a lot of small fry businesses to renegotiate contracts, pay for cloud products etc. and squeeze a lot of cash. On the other, it seems like the big fish, the blue chips aren't going to be held to ransom, and have the resources and financial will to move away from VMware and sue Broadcom in the process. And whatever cash Broadcom get in the short term is surely lost in reputational harm, lack of trust, lawsuits and lost customers in the future.
Grok is a general purpose AI and not a very good one. I'm sure it's great for parroting Elon's narrative du jour, but what company in their right mind would want to *pay* to use it when there are so many better alternatives out there which are designed for specialist and bespoke purposes?
WSL already provides syscalls to run user-land Linux over Windows, e.g Ubuntu. Containers are also user-land images making syscalls to the host so that part is more or less the same except there is a pseudo host sitting in the middle to set up cgroups, network and fs properly.
So presumably launching containers directly would be fast & lightweight since it wouldn't need to be done through a Linux dist. Even Docker for Desktop is using a small "docker-desktop" Linux dist to facilitate running images. Now that wouldn't be necessary. I could see it being useful in situations like build / CICD where you want to invoke some tool and be in and out without booting up an entire user-land for the sake of it.
I use it all the time to develop Linux apps while benefiting from the other software I use that is native to Windows. I even have docker containers and kubernetes going inside of it too. The purpose of the "native" containers is to forego the need to install Docker for Desktop which is actually commercial software (that uses WSL as its backend) and get slightly better performance than having to kick off Ubuntu to run a container when it can be run directly.
Having an agent doing stuff which is annoying and unwelcome is not only a waste of time but it costs the owner money in terms of tokens. What's the point?
Anyway the way to solve this is an explicit consent (e.g. .ai-consent.md) file in the project that lists what AI contributions are allowed and which are not. If there is no consent file then there is no consent. It should be that straightforward.
A long time ago Postgres was known as the "correct" SQL database with ACID compliance and MySQL was known as the "fast" database without ACID compliance. Correct or fast you can't have both. All the PHP scripts used MySQL because it was better at serving up pages fast.
But it turns out people really want ACID compliance so MySQL acquired a different backend storage engine called InnoDB which had it. But MySQL was confusing since now there are two engines boasting difference characteristics. Eventually the original ISAM engine got dumped but InnoDB won. But that was acquired by Oracle and then MySQL was acquired by Sun, and Sun acquired by Oracle and suddenly MySQL front and back is owed by Oracle. People had an issue with that and then MariaDB was forked from MySQL.
Meanwhile Postgres was chugging along being correct and getting more performant and new features with each release. I think the lack of drama and confusion has meant it has basically won between the two. Virtually every cloud service favours postgres, probably because they don't want Oracle lobbing some kind of sueball at them if they built off MySQL and the licence is also more permissive.
