Companies could just reduce the amount of tracking data they're trying to harvest - then they wouldn't need a banner. If you're annoyed then be mad at the company - not the law trying to offer you some way to protect your data.
> we employ these safety people to make sure our chat bot does not turn into Skynet
i think it's mostly about not showing up in some NYT article titled "look what crazy thing i got this AI to say". There were a bunch of those early on and it really hurt the cause. Microsoft had some famous ones, even prior to chatgpt, where the AI got pretty testy in the chat.
Most soldiers who disagreed with the Nazi left the Wehrmacht by 38 (or were pushed out, or were beaten and restrained for month in 'black site', often disaffected warehouses where socialists and Communist were 'retrained', basically proto-concentration camps).
In 42-43 some were reintegrated forcefully, and those I agree do not bear responsibility for the Wehrmacht war crimes.
My grandmother had a lot of stories about the Wehrmacht and was very disappointed that they were basically whitewashed from their pedophilic rapes, public executions and other group punishment.
I meant geniuses in the real sense, not colloquial sense.
If someone could reasonably expect a cushy 40 hour/week seven figure job, even a malicious personality wouldn’t risk criminal fraud without a much much bigger payout.
And to have dozens focusing on one company…
So anyone handling under a few hundred million per day are safe from that kind of coordinated attack.
For some sorts of "confusables", you don't even need Unicode in some cases. Depending on the cursed combination of font, kerning, rendering and display, `m` and `rn` are also very hard to distinguish.
For much of that time frame, the Soviets were also meddling in the Middle East. That the middle east conflicts were themselves part of the Cold War rather than something unrelated, is knowledge that has gone forgotten in the West, I think.
First, cookie banners are associated with a totally different legislation, not GDPR, and they began appearing long before GDPR existed.
Second, the EU is not to blame for cookie banners. Companies doing tracking via cookies are to blame. They always have the option to not have a cookie banner--just don't do the things that require cookie banners. They deliberately choose to do these things, and then people complain about the banners.
If there is a VC-backed for-profit company, the core part is how much value something brings. Words are cheap, so it is easy to add or remove them. Compare and contrast "Don't be evil" by Google.
"Safety" here works for both PR and hiring (a lot of talented engineers and researchers might flock to it), and maybe soft power for legislation.
I do not say that individual employees do not care about safety (well, a lot don't, what is very visible during this OpenClaw mania).
We process video, images, and documents through 20+ ML models simultaneously at Mixpeek. A single 10-minute video triggers transcription, visual embeddings, scene descriptions, face detection, object detection, brand safety classification, and more — all in parallel with different compute requirements.
We wrote up the full Ray architecture we use in production on KubeRay/GKE. Not a tutorial — more of a "here's what we actually run and what bit us."
Some highlights:
- *Custom resource isolation* — We use a synthetic `{"batch": 1}` resource to prevent batch pipeline tasks from starving Ray Serve inference replicas. Same cluster, zero interference, no runtime overhead.
- *Flexible actor pools* — Fixed-size `ActorPoolStrategy(size=8)` deadlocks when concurrent jobs compete for workers. `min_size=1, max_size=N` guarantees every job can make progress.
- *Shared preprocessing* — Naive approach runs S3 download + format normalization once per extractor. With 10 extractors on 1,000 files, that's 10,000 redundant reads. We preprocess once and fan out via Ray Dataset.
- *Distributed Qdrant writes* — Ray Data's `Datasink` API distributes vector DB writes across all workers with backpressure, instead of collecting everything on one node.
- *Fire-and-forget progress tracking* — A Ray actor as a shared counter lets workers report progress without blocking the pipeline.
- *Zero-CPU head node* — Learned this one the hard way when a runaway batch job took down our scheduler.
The post includes the KubeRay YAML, Ray Serve autoscaling configs, pipeline code, and the LocalStack parquet workaround that saved us hours of debugging silent hangs.
It is incredibly stupid and counterproductive to make this kind of statement publicly.
Most of the GAFAM companies are doing their utmost to try to reassure their European customers with a facade of sovereignty.
In all cultures, there is an expectation that you have to provide a name for yourself that is intelligible to the culture you're interacting with, both in written language and in speech. If your name is Albert and you are going to interact with many Japanese speakers, you'll have to call yourself アルバート in writing and pronounce your name as something like "Ah roo bay toe" to fit in. If you have a name whose pronunciation depends heavily on tones, such as a Mandarin or Vietnamese name, and you are going to interact with speakers of a non-tonal language, you'll have to come up with a version that you're happy with even if pronounced in the default neutral tone that those people will naturally use. If your name is 高山, you'll have to spell it as Takayama.
Similarly, if you're going to create an identifier for yourself that is supposed to be usable in an international context, you'll have to use the lowest common denominator that is acceptable in that context - and that happens to be a-zA-Z0-9. Why the Latin alphabet and numerals and not, say, Arabic, you might ask? Because Chinese and Indian and Arabic speakers are far more likely to be familiar with the Latin alphabet than with each other's writing systems.
At a certain point, if you want the people of your own country to have any sort of loyalty or deference for you, then you'll need to have loyalty or deference for them.
"But it's cheaper in our main geopolitical rival" doesn't quite wear like it used to.
In the context of AI research, there is no question that "existential" means "powerful AI literally kills every human being". It's a mainstream although not universal view among experts in the space that this is a serious possibility.
I bet it’s too late now. They will need very very persuasive arguments to kill all the initiatives, and while they may convince some governments and lobbying groups, I doubt they will manage to convince every IT responsible.
