[–] 2 points 17 hours ago

Mr Burke said the time between detecting a cyber vulnerability and exploiting it was previously about two and a half years, but with the advent of AI that window had narrowed to about 10 hours.

I assume somebody from Anthropic told him that, there's a reason "Patch Tuesday" was followed by "Exploit Wednesday".

  • source
  • [–] 5 points 1 day ago (1 child)

    That's only half the quote

    The US power sector would emit 2.1 billion metric tons of carbon dioxide by 2030, after factoring in planned AI projects, which is 618 million tons more than 2025 levels, or about a 42% increase.

    So the estimated total increase is 618 million tons

    The bulk of the expected emissions by 2030 would come from sources that are not directly linked to the grid, often built by tech companies to avoid the long wait for a connection. These could exceed 460 million tons, Fisher said, roughly the same as the annual emissions of Mexico.

    So over 460 million tons of the 618 estimate is primarily datacenters, nearly 3x of the rest of the country, which is only using 158 tons more.

  • source
  • parent
  • context
  • [–] 3 points 4 days ago (1 child)

    "Finding" is also rather generous, going by numbers from Anthropic and from the Linux kernel, things like Mythos are wrong 95% of the time.

    They're much easier to run though, compared to more reliable tools. So even being wrong 95% of the time wins out if you've got 100x more people running them.

  • source
  • parent
  • context
  • [–] 6 points 5 days ago

    So Scarlett Johansson was in their ads for a while back in 2014, made a bit of a stir since she was also an Oxfam ambassador, and Oxfam are against the settlements.

    When she had to choose between them, she chose Sodastream, and cut ties with the charity she had been involved with for more than 7 years.

    https://www.abc.net.au/news/2014-02-03/scarlett-johansson-quits-oxfam-amid-sodastream-israel-criticism/5235030

    Never really had a good opinion of her since.

  • source
  • parent
  • context
  • [–] 1 point 6 days ago (1 child)

    How does letting AI companies have total access to creators work without needing permission or licensing fees, help them exactly?

    The linked article is talking about protecting people making the models, not making the content (It actually specifically says that letting creators choose license their content would actually hurt them, and that AI companies should always have access)

  • source
  • parent
  • context
  • [–] 4 points 1 week ago (3 children)

    Disney will be able to negotiate and get a large chunck of money from the companies, any small time artist will be denied their art was used and not get anything.

    That's literally the current situation, that's why journalists and authors are trying to sue over it.

    The EFF is effectively siding with "Disney" here in this case

  • source
  • parent
  • context
  • [–] 1 point 1 week ago

    There's probably flaws that would allow executing attacker supplied code, but it's not like BIOSs had any notion of security (i.e. secure boot) to begin with, so you could just straight up have the system boot into your own image anyway.

  • source
  • parent
  • context
  • [–] 3 points 3 weeks ago (1 child)

    As nice as DANE is, it just shifts the burden of trust to your DNS provider. Like if you hop on Wi-Fi at a coffee shop, said network can simply MITM connections and supply their own certs via re-writing the DNS.

    You can encrypt DNS to prevent that, but now you need a way to ensure your DNS client already trusts the desired DNS server, etc.

    As much as people dislike CAs, they're a proven solution to a problem that's already implemented in nearly every device on the planet.

  • source
  • parent
  • context
  • view more: next ›