• 1 post
  • 1.7K comments
Joined 2 years ago
Cake day: February 10th, 2025
  • The CPU has a power saving mode that lets it slow down whenever a thread is sleeping (even a few milliseconds). In games, the main thread does short bursts of work then sleeps briefly. So the CPU keeps waking up “cold” at a low speed, causing occasional slow frames (the “1% lows”).

    The patch adds a window where the CPU won’t go into the power saving state immediately, so when the thread starts working again the CPU is at full speed already.

  • The Western model was doomed to failure from the start. The only barrier to entry was being able to download thousands of TBs of internet archives/books and to have a lot of compute.

    The math for these models isn’t proprietary and most CS students are exposed to machine learning and neural networks while in school.

    The only advantage western companies had was the ability to buy up the entire hardware market, pricing out domestic competition, and to use their politicians to manipulate trade policy in order to restrict sales of critical hardware to China.

    Every US tech company has dumped billions investing in an unsustainable business model with the hope of buying a global monopoly by strangling competition.

    China can destroy all of that by making their models open weight. The real money is in finding and implementing custom AI solutions… not in charging for access to the models. By having freely available models, they’re making the barrier of entry as low as possible.

    Not to mention that the insane amount of money being poured into hardware by US tech companies has created an environment where building fabs has a much shorter ROI, which also helps China’s development in that sector.

    US companies are playing Monopoly while China is playing Civilization.

  • Depends

    Most phones can be accessed by forensic software. The more secure phones are protected as long as they’ve just been restarted but before they’re unlocked for the first time.

    Some phones, like an up-to-date GrapheneOS on a supported Pixel, may resist forensic imaging after first unlocking but the information is in memory so there could potentially be zero days or other exotic methods of obtaining the data.

    A duress wipe lets you know 100% that the data is not accessible.

  • In addition, the law that he’s charged with relates to knowingly destroying property that the police have a warrant to sieze.

    There is no precedent of it being applied to border searches where they may not even have reasonable suspicion of a crime and can stop people for questioning on a whim.

    The State has absolutely no claim over your personal property at that point and destroying property that the police hope and dream of siezing isn’t the same as destroying property where they’ve demonstrated probable cause to a judge.

    This argument from the border patrol that you can be arrested for destroying data that is in your possession without ever been given due process and 4th amendment protections isn’t supported by precedence.

  • The profiles are independently encrypted with a key derived from the user’s profile PIN. If the other profiles are not logged in then their keys are not in RAM (and they get zeroed beforehand so they can’t be read out of unallocated memory).

    Even if the bad guys get root via a LPE in a dummy account there are no keys to find and the profiles’ storage are not mounted.

    You couldn’t trust the device after that however, who knows what kind of persistent spyware they could have installed.

  • You could say the same about drug cartels and most governments spend billions of dollars fighting them while outlawing their products and sanctioning countries/governments which allow them.

    I don’t expect the US Navy is going to be launching missiles at programmers in speedboats anytime soon, but it would be easy to sanction these companies and cut them off from the banking system.

    Except the government is run by people in the pockets of the companies and countries who are doing this.

  • I’m sure that case will be cited here.

    In similar cases, the court has found the company liable for using the books without paying for them (so they had to pay for every individual book) but the process of training an AI model was ‘sufficiently transformative’ (see: Authors Guild v Google) so as to not be considered a copyright violation.

    They may win damages equal to the costs of the books that were used, but they won’t get the models deleted or be able to claim damages for the models that were created.