Vendor-Lock-in bei Inferenzbeschleunigung? NIS2 macht die Rechnung neu. Proprietäre Blackboxes sind das neue Risiko.
#NIS2 #OpenSource
#NIS2 #OpenSource
Inferenzbeschleunigung: NIS2-Risiken proprietärer Stacks
Proprietäre Inferenzbeschleunigung tauscht Durchsatz gegen NIS2-Lieferkettenintransparenz und Vendor Lock-in. Offene Engines schließen die Lücke mit geringem
fluxhuman.com
November 14, 2026 at 9:10 AM
Vendor-Lock-in bei Inferenzbeschleunigung? NIS2 macht die Rechnung neu. Proprietäre Blackboxes sind das neue Risiko.
#NIS2 #OpenSource
#NIS2 #OpenSource
Fair point, I just remember about ultralights with registrations and I'm doubting they have blackboxes and stuff.
Notwithstanding unregulated unregistered ultralights though.
Notwithstanding unregulated unregistered ultralights though.
November 12, 2026 at 1:07 AM
Fair point, I just remember about ultralights with registrations and I'm doubting they have blackboxes and stuff.
Notwithstanding unregulated unregistered ultralights though.
Notwithstanding unregulated unregistered ultralights though.
I'm just kind of stunned this isn't like, the median lukewarm position! "Test the thing before you deploy to prod, and also don't trust weird closed source blackboxes with no prior reputation to stand on" just seems like the thing you should've always been doing anyway!
November 2, 2026 at 9:05 AM
I'm just kind of stunned this isn't like, the median lukewarm position! "Test the thing before you deploy to prod, and also don't trust weird closed source blackboxes with no prior reputation to stand on" just seems like the thing you should've always been doing anyway!
Wichtiger Unterschied: NP-Probleme sind keine Blackboxes. Schaut man sich das Zahlenschloss als Whitebox an, sieht man am Aufbau in der Regel ohne Ausprobieren, welche Kombination eingestellt ist.
November 2, 2026 at 10:25 AM
Wichtiger Unterschied: NP-Probleme sind keine Blackboxes. Schaut man sich das Zahlenschloss als Whitebox an, sieht man am Aufbau in der Regel ohne Ausprobieren, welche Kombination eingestellt ist.
LLMs are blackboxes-> you CAN’T know what you will get.
Two cents:
Can you trust it-> obvious answer NO -> extensive review + refactoring.
Are you competent and willing to maintain it properly-> rarely the case for must generated projects.
LLMs are niche tech forced to run the show.
Two cents:
Can you trust it-> obvious answer NO -> extensive review + refactoring.
Are you competent and willing to maintain it properly-> rarely the case for must generated projects.
LLMs are niche tech forced to run the show.
November 2, 2026 at 11:05 AM
LLMs are blackboxes-> you CAN’T know what you will get.
Two cents:
Can you trust it-> obvious answer NO -> extensive review + refactoring.
Are you competent and willing to maintain it properly-> rarely the case for must generated projects.
LLMs are niche tech forced to run the show.
Two cents:
Can you trust it-> obvious answer NO -> extensive review + refactoring.
Are you competent and willing to maintain it properly-> rarely the case for must generated projects.
LLMs are niche tech forced to run the show.
anyone else think subnautica 2 is the perfect opportunity to bring back time capsules? considering the whole mechanic of the colonists & players leaving blackboxes behind when they die, i feel like this could be a great chance to introduce a mechanic similar to the time capsules from SN1 somehow
October 30, 2026 at 12:08 AM
anyone else think subnautica 2 is the perfect opportunity to bring back time capsules? considering the whole mechanic of the colonists & players leaving blackboxes behind when they die, i feel like this could be a great chance to introduce a mechanic similar to the time capsules from SN1 somehow
@mandree.bsky.social sagt es: Wir müssen die digitalen Medienmonopole abschaffen!
Meta kann Parteien oder Kandidaten vor einer Wahl jederzeit in ihrer Sichtbarkeit einschränken, es ist aber extrem schwierig, das zu beweisen. Die Algorithmen sind Blackboxes.
Meta kann Parteien oder Kandidaten vor einer Wahl jederzeit in ihrer Sichtbarkeit einschränken, es ist aber extrem schwierig, das zu beweisen. Die Algorithmen sind Blackboxes.
Vorwürfe gegen Facebook: Untergräbt Meta die Demokratie? - Interview mit Martin Andree
Vorwürfe gegen Facebook: Untergräbt Meta die Demokratie? Interview mit dem Medienwissenschaftler Martin Andree, Uni Köln.
www.deutschlandfunk.de
October 28, 2026 at 11:39 AM
@mandree.bsky.social sagt es: Wir müssen die digitalen Medienmonopole abschaffen!
Meta kann Parteien oder Kandidaten vor einer Wahl jederzeit in ihrer Sichtbarkeit einschränken, es ist aber extrem schwierig, das zu beweisen. Die Algorithmen sind Blackboxes.
Meta kann Parteien oder Kandidaten vor einer Wahl jederzeit in ihrer Sichtbarkeit einschränken, es ist aber extrem schwierig, das zu beweisen. Die Algorithmen sind Blackboxes.
Banks Beware Agentic AI Blackboxes in Modernisation Plans
my.sociabble.com
October 25, 2026 at 9:54 AM
Perhaps with loads, loads and LOADS of blackboxes, you could talk about how Legend Of The Overfiend changed the west's perception of anime for the better and for worst, especially with how certain people have used it to write off the entire anime medium as "pornography", like you know who, and...
October 25, 2026 at 12:13 AM
Perhaps with loads, loads and LOADS of blackboxes, you could talk about how Legend Of The Overfiend changed the west's perception of anime for the better and for worst, especially with how certain people have used it to write off the entire anime medium as "pornography", like you know who, and...
Never install vibecoded mods and tools, no matter how "good enough" they seem
Absolute blackboxes of "who knows what actually does"
Absolute blackboxes of "who knows what actually does"
October 19, 2026 at 9:45 AM
Never install vibecoded mods and tools, no matter how "good enough" they seem
Absolute blackboxes of "who knows what actually does"
Absolute blackboxes of "who knows what actually does"
Und? Wie schließt Du an der Stelle Diskriminierung aus? Durch Wahrscheinlichkeit?
Durch sogenannte KI? Durch Überwachung? Durch Überwachung sogenannter KI? Durch Blackboxes?
Wie soll denn da der Mensch nicht mehr das schwächste Glied sein? Wie die Qualität erreicht sein, von der Du sprichst?
Durch sogenannte KI? Durch Überwachung? Durch Überwachung sogenannter KI? Durch Blackboxes?
Wie soll denn da der Mensch nicht mehr das schwächste Glied sein? Wie die Qualität erreicht sein, von der Du sprichst?
October 17, 2026 at 9:24 AM
Und? Wie schließt Du an der Stelle Diskriminierung aus? Durch Wahrscheinlichkeit?
Durch sogenannte KI? Durch Überwachung? Durch Überwachung sogenannter KI? Durch Blackboxes?
Wie soll denn da der Mensch nicht mehr das schwächste Glied sein? Wie die Qualität erreicht sein, von der Du sprichst?
Durch sogenannte KI? Durch Überwachung? Durch Überwachung sogenannter KI? Durch Blackboxes?
Wie soll denn da der Mensch nicht mehr das schwächste Glied sein? Wie die Qualität erreicht sein, von der Du sprichst?
With the way things are going right now and everyone creating code blackboxes and no proper security measures, I worry that we'll actually approach some .hack universe level catastrophic cyber incident, that will crash the entire digital infrastructure, worldwide.
October 15, 2026 at 1:10 PM
With the way things are going right now and everyone creating code blackboxes and no proper security measures, I worry that we'll actually approach some .hack universe level catastrophic cyber incident, that will crash the entire digital infrastructure, worldwide.
Satya Nadella says we should assume all AI models are ‘compromised’ #Technology #AI #GenerativeAI #satyanadella #aimodels #aiethics #microsoft #blackboxes
https://puretech.news/read?id=234805661286793216
https://puretech.news/read?id=234805661286793216
Satya Nadella says we should assume all AI models are ‘compromised’
Microsoft’s CEO calls for putting an ‘emergency brake’ on AI.
puretech.news
October 11, 2026 at 12:45 AM
Satya Nadella says we should assume all AI models are ‘compromised’ #Technology #AI #GenerativeAI #satyanadella #aimodels #aiethics #microsoft #blackboxes
https://puretech.news/read?id=234805661286793216
https://puretech.news/read?id=234805661286793216
"If neural networks—often criticized as opaque blackboxes—have their greatest interpretive difficulties when the underlying biology would most benefit from their ability to represent complexity, how can we expect DCMs [deep comparative methods] to possibly improve the situation?"
September 29, 2026 at 2:27 PM
"If neural networks—often criticized as opaque blackboxes—have their greatest interpretive difficulties when the underlying biology would most benefit from their ability to represent complexity, how can we expect DCMs [deep comparative methods] to possibly improve the situation?"
We aren't programming #AI; we’re growing it. #DeepLearningModels are vast #blackboxes whose internal logic can't be read like standard code. Engineers don't know what a frontier model will actually do until they run it & observe results. The public needs to understand this reality. #BeAIAware 1/2
September 26, 2026 at 6:28 PM
We aren't programming #AI; we’re growing it. #DeepLearningModels are vast #blackboxes whose internal logic can't be read like standard code. Engineers don't know what a frontier model will actually do until they run it & observe results. The public needs to understand this reality. #BeAIAware 1/2
None of these tech geniusses have the slightest idea on why their AI blackboxes do half of the things they do, yet we are somehow to trust them. Yeah. Sure.
September 24, 2026 at 5:41 PM
None of these tech geniusses have the slightest idea on why their AI blackboxes do half of the things they do, yet we are somehow to trust them. Yeah. Sure.
huge blackboxes that just gonna work
September 14, 2026 at 6:56 AM
huge blackboxes that just gonna work
“yes, I’m prepared for linear algebra blackboxes to take my job, and actually that’s a good thing.”
daily positive affirmations fit for the 21st century.
daily positive affirmations fit for the 21st century.
September 10, 2026 at 10:18 PM
“yes, I’m prepared for linear algebra blackboxes to take my job, and actually that’s a good thing.”
daily positive affirmations fit for the 21st century.
daily positive affirmations fit for the 21st century.
Bias in, bias out. The billionaires are desperate calcify heirarchies of white supremacy & patriarchy into their blackboxes to control world's collective digital decision makers and knowledge libraries. They are forcing it on us before the world becomes an actually enlightened or empathetic place.
AI hiring tools can reject qualified people "for reasons no one—including the employer—can adequately explain," warns NYU School of Law senior affiliated research scholar and former EEOC chair Charlotte A. Burrows. She says gutting disparate impact law makes it worse.
A Labor Day Prescription for Preserving Fair Employment in the Digital Age
We can’t let technological progress turn back the clock on civil rights, writes Charlotte A. Burrows.
techpolicy.press
September 3, 2026 at 11:37 AM
Bias in, bias out. The billionaires are desperate calcify heirarchies of white supremacy & patriarchy into their blackboxes to control world's collective digital decision makers and knowledge libraries. They are forcing it on us before the world becomes an actually enlightened or empathetic place.
Polizeiliche Datenanalyse und menschenrechtliche Grenzen: Hauptsache automatisiert? Ob „Blackboxes“ wie #Palantir genutzt werden, will die Regierung den nutzenden Behörden überlassen verfassungsblog.de/palantir-dis...
Wegducken vor den Fragen des digitalen Zeitalters: Die automatisierte Datenanalyse braucht menschenrechtliche Grenzen, weil menschliche Grenzen obsolet werden
In mehreren Bundesländern arbeitet die Polizei bereits mit automatisierter Datenanalyse. Das soll nun auch für das BKA und die Bundespolizei möglich werden. Der Gefahr, dass Beamt*innen auf der Suche…
verfassungsblog.de
August 31, 2026 at 7:53 PM
Polizeiliche Datenanalyse und menschenrechtliche Grenzen: Hauptsache automatisiert? Ob „Blackboxes“ wie #Palantir genutzt werden, will die Regierung den nutzenden Behörden überlassen verfassungsblog.de/palantir-dis...
they do sweeps using various ML blackboxes and human moderator sweeps.
Tho rationale is mostly preventing creators from coaling users into other, better platforms
Anyways if your target audience is opportunistic pedo who cannae inta TOR browser (dumb) you want a platform where contact leaves as>
Tho rationale is mostly preventing creators from coaling users into other, better platforms
Anyways if your target audience is opportunistic pedo who cannae inta TOR browser (dumb) you want a platform where contact leaves as>
August 28, 2026 at 11:37 PM
they do sweeps using various ML blackboxes and human moderator sweeps.
Tho rationale is mostly preventing creators from coaling users into other, better platforms
Anyways if your target audience is opportunistic pedo who cannae inta TOR browser (dumb) you want a platform where contact leaves as>
Tho rationale is mostly preventing creators from coaling users into other, better platforms
Anyways if your target audience is opportunistic pedo who cannae inta TOR browser (dumb) you want a platform where contact leaves as>
changes and downtime inducing bugs at the same time when you need a new feature. Configs and things are in git. Avoid blackboxes if you can this results in using a lot of OSS. EAP-TLS over any other solution, everything else is more config complex and less secure.
August 24, 2026 at 1:36 AM
changes and downtime inducing bugs at the same time when you need a new feature. Configs and things are in git. Avoid blackboxes if you can this results in using a lot of OSS. EAP-TLS over any other solution, everything else is more config complex and less secure.
... unless you mean neural networks which are one-off blackboxes that... you can add guardrails onto.
Specializing too much in one area pulls away from accuracy in another (see: overtraining), being less than 100% precise is by design. (2/3)
Specializing too much in one area pulls away from accuracy in another (see: overtraining), being less than 100% precise is by design. (2/3)
September 11, 2025 at 11:25 PM
... unless you mean neural networks which are one-off blackboxes that... you can add guardrails onto.
Specializing too much in one area pulls away from accuracy in another (see: overtraining), being less than 100% precise is by design. (2/3)
Specializing too much in one area pulls away from accuracy in another (see: overtraining), being less than 100% precise is by design. (2/3)