Search results

  1. P

    Uber’s mandatory arbitration upheld in case over severe crash injuries

    It sounds like there's legitimate question that the victim actually waived the rights, and not another party with access to there phone. And indeed, I'd question whether that child even read the 150 page document, never mind consulted with there attorney! I'm not terribly hopeful the courts...
  2. P

    OpenAI plans tectonic shift from nonprofit to for-profit, giving Altman equity

    The cap on investor return is irrelevant, it's 100x and doubles every 4 years. If they're trying to raise a $150B round, it means investors have to wait until 2029 to receive returns in excess of a $15T valuation. Nobody is going to say "oh, only $15T? Not worth my time". If you have the...
  3. P

    Due to AI fakes, the “deep doubt” era is here

    The truth is often discoverable. The people in Kamala's crowds do or do not exist. They can be interviewed. They have memories of where the crowd extended to. How closely packed people were. Even if the crowd is of partisans, lies about the crowd size would vary where the truth should be more...
  4. P

    As contentious California AI safety bill passes, critics push governor for veto

    So, from like 1950 (Turing's Computing Machinery and Intelligence) - mid 2010's (ImageNet results, BERT release) a ton of AI/NLP work was done in universities. There were certainly corporate efforts, but the academics could play along. Then we had a decade where all the major stuff was...
  5. P

    As contentious California AI safety bill passes, critics push governor for veto

    Meta has been freely releasing its model weights. There's no sucking bytes back over the wires. If the original developer is responsible for modifications of its weights, Meta becomes responsible for the models it releases publicly and that stops. Suddenly there's a $100M barrier to entry to...
  6. P

    DeepMind AI rivals the world’s smartest high schoolers at geometry

    Oh leave the poor undergrads be, they're trying their best.
  7. P

    Turning down holiday invites isn’t going to obliterate your social life

    I think they meant initiate social interactions, not ask if events are occurring. If a friend frequently declines invitations to do something, but is actively inviting me to things as well, it feels like we're both just busy. If they decline my invitations and don't invite me to anything, the...
  8. P

    Study finds no “smoking gun” for mental health issues due to Internet usage

    It's really hard to run randomized experiments with social media or internet access (especially blinded studies!). There's also all sorts of weird counter-effects if you tried - maybe a teen cohort would be happier without Instagram, but if you randomly prevent a handful from using it, they...
  9. P

    Report: OpenAI tried—and failed—to hire Anthropic CEO to replace Sam Altman

    The implementation details are tricky, but why wouldn't a free floating brain be intelligent? Between shut-in syndrome and those with spinal injuries there are humans who aren't all that much more than that, and I'd say they are absolutely still as fully intelligent as the rest of us. I'd set...
  10. P

    OpenAI employees revolt after board names new CEO and Altman heads to Microsoft

    Oh sure, and 5/7 just signed the letter urging the board to resign. My point was that the decision makers had no stake in being friendly towards an exit. The angry employees certainly did.
  11. P

    OpenAI employees revolt after board names new CEO and Altman heads to Microsoft

    Isn't the oldest member of the board left like 40? I get that tech is a young man's game, but is that really the boundary for old now?
  12. P

    OpenAI employees revolt after board names new CEO and Altman heads to Microsoft

    That's why the board doesn't hold equity.
  13. P

    OpenAI employees revolt after board names new CEO and Altman heads to Microsoft

    I think that's fine, though? If the board is right - and I kind of think they are - the for-profit motive was controlling the work. OpenAI went from "If you're an academic researcher maybe we'll give you access to GPT-2", to "insert credit card here" on GPT-4. The whole point of OpenAI was to...
  14. P

    Deepfake nudes of high schoolers spark police probe in NJ

    And look, if we get there as a society, where everyone is constantly just walking around naked and nobody cares, then sure, this would be less of an issue. But it's preposterous to try to use victims as a wedge to change culture. First we get to the place where nobody cares, then we can stop...
  15. P

    University of Chicago researchers seek to “poison” AI art generators with Nightshade

    The company the released a self driving car without testing whether it knew the difference between a person and a cat? And programmed it to actively run over cats? With our without adversarial images in the wild, if you don't vet whether your system consistently avoids killing people that's...
  16. P

    University of Chicago researchers seek to “poison” AI art generators with Nightshade

    Eh, I think this is where the field will have to go anyways. You can accomplish the same effect just labeling dogs as cats and waiting for the photos to be scraped: that's already being done by artists. The LLM field has put substantial work into cleaning the atrocious quality you get blindly...
  17. P

    Universal Music sues AI start-up Anthropic for scraping song lyrics

    On the contrary, see the rendition of "I will survive" above. Even some of the most recognizable prose: "it was the best of times, it was the worse of times" - would we collectively have any interest and memory to write the following lines as banter? "it was the spring of...something? fall of...
  18. P

    So far, AI hasn’t been profitable for Big Tech

    When a deep learning person says "probabilities", they usually mean "I ran softmax on this layer so now the values add up to 1." It's complex in that there are billions of parameters, but all you're really doing is a little "add and multiply" dance over and over again. In about 3 years, we got...
  19. P

    OpenAI may jump into AI hardware amid high costs, supply constraints

    My first instinct was "it's a couple seconds on a beefy GPU, it can't possibly be that much energy. But it appears I had no idea how little energy charging a cellphone takes. It's a different extreme, but back of the envelope you can charge your cellphone a good 200 times for every load you run...
  20. P

    AI firms working on “constitutions” to keep AI from spewing toxic content

    No! There is subjectivity on the margins, and different approaches to ethical reasoning (utilitarianism, virtue ethics...) sometimes come to conflicting conclusions, but reasonable ethical systems agree on most situations. "Should I squish this or that guy on a trolley" is an interesting thought...