Grew up with an Amiga 500. Studied CS (Compilers and OSs). Did IoT and sensor networks for my PhD. Stayed at the university to now do data science and AI stuff lik everyone else.
This is the reason I switched probably mor than a year or two ago. YouTube remains a cat and mouse game. Strange things with those forks is that they always add some more streaming sources Mike BkliBili but keep the old stuft. Actually soundcloud only works very limited lately.
Wouldnt call it regulatory blood lust at the moment. I think the competitive situation that is however solely focussed in performance at any cost and not at compliance at all is forcing AI labs to play with fire. The risk for them is an even bigger regulatory backlash. Totally different industry : Chinese ebike manufacturers are currently pushing the rules for motor strength to the max due to competition. What will likely happen is stricter regulation of max support and motor power if there are incidents. Particularly there will be more regulatory fragmentation with little common denominator. This will hurt the whole industry's growth. In this industry it kind of still works as an agreement of non-Chinese manufacturers.
Interestingly Russia I think itself presented its 2012 law as modeled on the American Foreign Agents Registration Act at the time. A bit worrying is that the US DoJ starts to create also a bit of the same chilling effect speaking about “Liability for Unregistered Foreign Agents". Don't get me wrong: there is huge differences in freedom of speech and I do not say this is the beginning of the end. But it seems at least intellectually interesting to draw comparisons.
Can someone explain to me how such self-awareness can be forced into the model. I mean I guess the pre training data could contain all sorts of stuff. How reliable are those hacks. I know that a lot of open weight models answer that they are Claude in the absence of a system prompt. I find destillation not that much of a plausible explanation as typically claude would probably not mention that it is Claude all the time. I find it rather plausible that a foreig. system prompt made it into pre-training. But again: I have no clue how much care is given by models to leave traces for destillation (for closed weights) or post training (for open weights).
Cool, I actually applied for a diploma thesis for doing the PIC 16 port around 2004. Raphael Neider actually was quicker then me, so he maintained the port for our lab at the Uni Karlsruhe (i am still in the same lab). We needed the port because was needed to allow people to compile our OS for our sensor nodes at the time. Still did a lot of testing of the PIC 16 port at the time because my thesis was about implementing a filesystem and a REST like API for the nodes.
I am not saying that what they say is wrong. However, this is about current phones. If your goal is supporting aftermarket phones the case looks very different. It is the only way to get a decent level of security. The problem is IMHO how graphene communicates. It is mixing tons of different things to always make their communication more 'bold'. Then they complain about people complaining about their communication style. It is my personal choice, but I accept less security while not having to care about GrapheneOS announcements. I value their work in a similar way that I value WikiLeaks.
A lot of neural engine, particularly in the embedded domain (ARM/RISC MCUs) have the same problem. Designing other models means on top of this means a lot of profiling to get convolution blocks right to get good speedups. (We optimized this in the past e.g. using Neural Architecture Search on super networks)
The core criticism to the vendor is the missing security element. This is a fair point.
However, calling out on e/OS in the same post, seems to me very counter-productive. We need more OS vendors and less infighting. Calling all custom ROMs insecure and claiming to be the only one is IMHO 'drama'. Particular there are contributions of me microG that are helpful if you want to de-Google ones phone. Graphene has a different approach: fair. People will use GrapheneOS if they share their goals.
Fairphone is about sustainability and a bit about not supporting major tech like Google. I don't think this hurts. In an ideal world we could have both. But sustainability seems to be a non-goal of GrapheneOS.
One of the reasons I haven't switched to Graphene is this kind of drama. I wish there was more middle ground. I don't think the post is in any way constructive and will lead to fairphone to consider adding a secure element.
While many things might be legal (or hasn't been ruled clearly illegal yet) /under different jurisdictions, we are still in the process of figuring out what we accept as ethical. As the output of models can't be easily copyrighted, destillation is equally disputed. Particularly if the primary model interaction was not destillation (as in this case) IMHO it will be legally quite difficult to restrict secondary use for training. In the end we have to find a legislation and probably even international treaties that account for the fact that classical copyright is beginning to become an obsolete concept.
This is true, as preprints are addressing the 'problem' at a later point in time. I guess git is better if you do not want to assign an identifier to an immutable version yet.
I like it a bit for the inspiration what could make sense, however, as there is little ablation studies, it is really difficult to understand the effect of most of the features.
90% of my apps (all except for banking, streaming and ev charging app) are installed without play store. Beyond FDroid with custom repos, I actually came to like Komi-Store in combination with obtainium. I really wonder what will happen when Google enforces their own registered developer signatures. I hope there is a crucial mass of apps outside of Play store until then to generate some heard protests.
I don't think it is too much power or even the wish for more. It is IMHO more the disconnect of EU beaurocrats and the Brussels bubble. We need more direct legitimation given the amount of power they have (actually down to the DGs).
Governments are real customers. The German government, I guess, pays about 30 times more annually to Microsoft than the budget of the STF. The other component of sovereign tech is that governments are actually using FOSS. There is some room there for improvement.
I think the biggest value of choosing gitlab is the social network. I think federation is the right way, but it seems a mastodon vs linkedin problem. Would be good if there would be bridges between all those fediverses. A nice one is also ngit [0]
In Germany it is assumed illegal without prior consent or with calling machines [0]. However, practically you get a few of those calls per week. Also often from international numbers.
People must realize that it is easy to built some sentiment against on part of legislation that seems to restrict them, however, it is much more difficult to establish legislation that serves everyone. In the end we need smart people in politics and lawmaking instead of populists and libertarian cry-babies.
I was actually surprised by the selection of cities in the OP until I learned that the author is the director of AI at der Spiegel. Related to the other comments on the 3D: maybe this can evolve to some gaussian splatting/diffusion world model thing...
I wonder what the 'stupid EU law' is in this case. EU allows to use copyrighted materials for data mining. This in theory created a more stable legal basis than 'fair use'. Now a court ruled on its limits. In the US there were similar law suits on using copyrighted work for training. I guess the state of copyright is complicated these days. Germany's economy might fold for other reasons but not because of court rulings that might create a stable legal framework.
We should make sure that the indirect cost for CO2 are added to the price of energy and imported products depending on the mix. It is one of the few areas where I think a market can fix things.
Google scholar has become the way to search papers (which is somewhat worrisome). What people want from there is a download for all the stuff that does not list an author copy. Open access IMHO is just a reaction to the fact that mostly you would not need a subscription anyhow. Now the authors are paying upfront or universities are paying flat for all their researchers. The problem is now the incentives are not increasing the number of subscriptions but increasing the number of papers published.
This is really an interesting case. Hope to see a ruling here. The edge cases are really interesting as well, like the fake pin on the back of the phone. I also wonder generally about the 'destruction' of data
Wouldn't the government need to prove that there is no backup, because just making it more difficult (like hiding) would probably not call for the paragraph. Unfortunately for the rest of us the defence is based on more proven grounds. I guess only plausible deniability is helpful: I e.g. would love to see my trusted android space being empty/recreated on false password.