I already asked on another message board too, but:
Can someone tell me how this technically can happen?
I assume HuggingFace performs benchmark testing using containerized versions of the LLMs, or what do they mean by sandbox? So the model was able to 'escape' the container? I'm not following here.
Also, is this an incredible feat or just a lucky find (stolen credentials)?
It is the OAI ExploitGym agents (on GPT 5.6-Sol with guardrails turned off) that escaped the sandbox, found a zero day in HF production dataset and exploited it.
Biggest thing I miss about all of this was the gatekeeping and curation really. There's so much garbage on Spotify, it's hard to find good music, and the recommendations I get always safely stay in what I already listened to.
The way around this is the people that listened to this also listened to that feature. That's the main way I discover new music these days. I find something new that I like and then I explore whatever it is that people listen to that also listened to that song. If you don't do that, the recommendations are basically more of the same shit. All the B-singles of an artist you like some songs of. Or worse: "you're old, here's some old music for you!". Or even worse, "your ip address is in Germany! You know you are a closet German! Here's some German hoompahpah music for you!".
Amazon and Youtube are equally useless when it comes to recommendations. All the machine learning talent in the world and they are utterly useless. I clicked a young ones clip on youtube a few weeks ago. Now my recommendations are 40% more f*ing young ones clips.
But randomly clicking stuff on Spotify reminds me of 25 years ago where you'd randomly download some shit and then listen to it. I also miss the art of a well produced album. I can't listen to individual songs of a good album. I have to play it beginning to end. I hate all the bonus tracks that Spotify slaps on albums. The whole point of the last track on Dark side of the Moon is the fade out to silence at the end. But that's just me.
I wish they would stop breaking my playlists by randomly breaking links to songs when they get replaced or re-imported. Seems I have to hunt down replacement tracks for 10-15% of my carefully curated playlists every year or so. Usually they are still there. But in some cases entire albums disappear. All the endless remasters, best off collections, etc. that they keep churning out result in endless breakage. How hard can it be to automatically replace those songs with the exact same recording on a different album?!
Fwiw, it's annoying, but you can go into your YouTube watch history and delete that one video from it, which should stop the recommender from suggesting those going forwards.
One thing I like to do on Spotify is to make a "Blend" playlist with a friend whose music taste overlaps with mine slightly. Not for its intended use of creating a playlist for you both to listen to at the same time, but instead I find it's useful at pulling your recommendations out of the safe zone (and towards the taste of your friend). Sounds a little bit dystopian that I'm cutting my friends out of the friend-recommendation part of music discovery, but if I find a new artist I like I usually make sure to mention it to the friend.
It was really wild, out of the box you would get: sound themes, color themes, icon themes, startup sounds, wallpaper themes, different loading animations, am I forgetting something? :-D
Custom css email themes aka "stationary", which other people can see on emails from you. These are still a very prominent part of email but I don't see anyone expressing their individuality with it anymore. It feels like a huge blast from the past to get one with any kind of theme that doesn't look like something a bank would use.
There was also "smileys" but I never used those because smileys were considered tacky. Yet it's the thing that stuck harder than anything else.
Yes, except it was much more colorful and vibrant. We had pixel art instead of the the deliberately-monochromatic committee-directed stick-figure horrors of today.
I will die on the hill that Java was a good language, and had the potential to leapfrog us from where we are by at least a decade.
But it got hobbled by the awful, awful enterprise style culture, cultural misunderstanding of OOP (especially inheritance), and corporation shenanigans (fucking oracle).
I think this is a selection bias speaking rather than a reasonable reflection of what goes on in the Java world. Some insanely sophisticated and high quality technologies are written with Java.
The problem is like with JS or PHP, it is ubiquitous in many settings. There are a lot of people who can use it because it was the default language taught in CS programs, many corporate settings for decades, or similar. It’s the runtime for android devices. It’s everywhere. Of course you’ll encounter a lot of low quality developers.
Your comment mostly indicates that you haven’t been fortunate enough to encounter the high quality Java devs, not that they don’t exist. They exist and they build world class software that backs massive systems like elastic search, Kafka, spark, or Cassandra.
When I started to use elastic search I found out that with some incorrect queries you could "poison" the process entirely and it would respond incorrectly to every single query from that moment on, until you killed it and restarted it.
They responded to my issue several years later. I had changed jobs and I couldn't care less any longer.
Compose and AOT compiled binaries would be amazing (GraalVM Native Image kinda thing) but it doesn't look very easy at the moment. Leyden with a regular JVM might be the best we get.
Compose UI apps can be compiled to native binaries already, via Kotlin's LLVM backend, though at the moment only the macOS/iOS targets have proper (official) support. Last I looked (a few years ago now), the Linux and Windows targets shouldn't be too far off, since it's all built on top of Skia already, someone just needs to care enough to put in the work. (But since right now you already get coverage for all platforms between JVM and Wasm, not to mention hot reload support on the JVM, there's little motivation to do so.)
We were writing and shipping desktop applications with it back in the nineties. Although many of the arguments against it were similar to the arguments against Electron today.
I think the UI look and feel was very ugly for many users and that caused its demise. The cross platform skin was ugly. The native skins were in the uncanny valley.
The framework was reasonably good for its time. By the time good looking UI frameworks came, the bad reputation was already set.
Even the later JavaFX was a tasteless exercise. I opened some apps and you could tell within 1 second that something was wrong because all the text was using fugly non-platform-native (or somehow screwed up) text rendering.
reply