unkz's recent activity

  1. Comment on Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers in ~tech

    unkz
    Link
    Sounds like Google would have to basically pull out of that market or be faced with absurd liability? Or at least all AI responses.

    Sounds like Google would have to basically pull out of that market or be faced with absurd liability? Or at least all AI responses.

  2. Comment on Claude Fable 5 and Claude Mythos 5 in ~tech

    unkz
    Link Parent
    Possibly try an anonymous session so it doesn’t fetch your memory?

    Possibly try an anonymous session so it doesn’t fetch your memory?

  3. Comment on Are there any games that had their development abandoned that you followed where you wish that continued/completed development? in ~games

    unkz
    Link
    I wish we got more sequels to all the old Sierra games. Space quest, quest for glory, king’s quest … all lots of fun. I know space quest 7 in particular was deep in development when they shut down.

    I wish we got more sequels to all the old Sierra games. Space quest, quest for glory, king’s quest … all lots of fun. I know space quest 7 in particular was deep in development when they shut down.

    1 vote

  4. Comment on 'Chill, relax, we can’t control everything' - relive Gianni Infantino news conference on eve of World Cup in ~sports.football

    unkz
    Link
    Really what should have happened. America should be shut out of world events as long as they keep disparaging and disrupting the world. Anyone could have predicted these outcomes given their awful...

    Without his engagement and involvement, I think it would have been, simple as that, impossible to organise a World Cup in the United States.

    Really what should have happened. America should be shut out of world events as long as they keep disparaging and disrupting the world. Anyone could have predicted these outcomes given their awful and highly public positions on visitors from other countries.

    6 votes

  5. Comment on Claude Fable 5 and Claude Mythos 5 in ~tech

    unkz
    Link
    The safety filter clearly needs a ton of refinement. I’ve been blocked several times now for very innocuous code. Some people have also been complaining that the contents of their memory have...

    The safety filter clearly needs a ton of refinement. I’ve been blocked several times now for very innocuous code. Some people have also been complaining that the contents of their memory have basically locked them out of using Fable at all, because they work in chemistry. Also, there was a period in time today when their safety filter was simply offline and I couldn’t use fable at all.

    2 votes

  6. Comment on How Terry Tao became an evangelist for AI in math in ~science

    unkz
    Link Parent
    Well, not entirely. A big part of this is being able to use LLM-type AI to generate testable Lean code to grind through the grunt work.

    Well, not entirely. A big part of this is being able to use LLM-type AI to generate testable Lean code to grind through the grunt work.

    Yet Tao had a way forward in mind. He didn’t think AI would replace human mathematicians anytime soon, but he did consider it particularly well suited to helping solve certain types of complex mathematical problems: ones that could be broken into thousands of small, manageable subproblems — essentially the same class of problems that worked well for Polymath projects. At that scale, mathematicians could employ AI to solve large swaths of the easiest subproblems, with its results outputted as formal proofs that Lean could check, and step in to handle the most difficult remaining questions themselves. In 2024, Tao was promoting this vision to anyone who would listen, and following the PFR project, he had realized that if he really believed in the work, he needed to step up and lead it himself. He also knew right away which problem he would start with.

    4 votes

  7. Comment on How Terry Tao became an evangelist for AI in math in ~science

    unkz
    (edited )
    Link
    Anything Tao says is probably worth reading, but this is very interesting stuff. The intersection of Lean, LLMs, and crowdsourcing has massive potential. Here’s the state of Equational Theories by...

    Anything Tao says is probably worth reading, but this is very interesting stuff. The intersection of Lean, LLMs, and crowdsourcing has massive potential.

    Here’s the state of Equational Theories by the way.

    https://teorth.github.io/equational_theories/dashboard/

    The article seems to say that it ground to a halt with 4 hard cases remaining. From that page it seems that they solved 2 more, leaving only 2 unproven cases remaining.

    3 votes

  9. Comment on Tildes Survey #8: What is your favorite video game? in ~talk

    unkz
    Link
    I kind of have eras that don’t really overlap. Like my favourite childhood game was final fantasy 1, which I don’t know how you would even compare with fallout 4, which is probably my favourite...

    I kind of have eras that don’t really overlap. Like my favourite childhood game was final fantasy 1, which I don’t know how you would even compare with fallout 4, which is probably my favourite game of adulthood.

    6 votes

  10. Comment on rsync and outrage in ~comp

    unkz
    Link
    If I understand this correctly, the claude contributions were limited to rewriting the test suite, and did not actually introduce bugs in the system?

    If I understand this correctly, the claude contributions were limited to rewriting the test suite, and did not actually introduce bugs in the system?

    5 votes

  11. Comment on What do you think of robots in the military? in ~tech

    unkz
    Link
    This is an interesting question to me, and I think it’s especially interesting on the broader scope of automation. The concern you have with automated kills is, I assume, the chance of collateral...

    Should robots be remotely controlled at all times or should they be automated?

    This is an interesting question to me, and I think it’s especially interesting on the broader scope of automation. The concern you have with automated kills is, I assume, the chance of collateral damage — civilians being killed by autonomous agents. However, human operators kill civilians all the time. So to me, it seems like the ethical line has a few components to it.

    • accuracy relative to human operators in killing desired targets versus undesired targets is the obvious one since it feels very apples to apples.
    • the other though, a little more difficult to quantify and that is how many innocent lives are being lost relative to invader kills. Even if collateral damage were higher on the side of Ukraine for instance, if this results in a higher total kill count of invaders than could be achieved by solely using human operators, I could still consider this to be ethical since killing a killer saves additional lives — all the soldiers and civilians that that invader would have killed. What that level of additional collateral damage is, I couldn’t give a firm answer to though.

    The more interesting question to me is at what point does it become unethical to use human operators at all? We may be heading towards that zone with self driving cars, for example.

    1 vote

  12. Comment on What do you think of robots in the military? in ~tech

    unkz
    Link Parent
    My understanding is we are basically there now. On board targeting for loitering munitions is in use today. https://www.bbc.com/news/articles/cdjp0n7rn41o

    My understanding is we are basically there now. On board targeting for loitering munitions is in use today.

    https://www.bbc.com/news/articles/cdjp0n7rn41o

    Ukraine's Hornet drones are equipped with an AI-targeting system which has been trained on thousands of hours of videos of Russian military targets gathered over the last four years, Nick Brown, a weapons expert from defence intelligence company Janes, told BBC Verify. They can also access the Starlink satellite network to connect to operators over longer distances, a system that is also more resistant to jamming by Russian forces.

    2 votes

  13. Comment on The fall of the theorem economy in ~science

    unkz
    Link
    I think one interesting little thread that hasn’t been explored a lot is the actual capacity of human understanding. What I’m really interested in is theorems that may now be provable yet cannot,...

    I think one interesting little thread that hasn’t been explored a lot is the actual capacity of human understanding.

    The likely outcome is that formalized mathematics will now develop in two separate layers, an intelligible layer embodied by Mathlib, and an unintelligible layer we might call Mathslop, a library of results that are known to be correct via proofs that no human has ever understood.

    What I’m really interested in is theorems that may now be provable yet cannot, because of human absolute limitations, ever be completely understood by humans. Even when expressed in the most parsimonious form, the complexity may exceed human understanding. Yet, these incomprehensible proofs may yet resolve as simple statements that are themselves usable as elements of reasoning — perhaps that will be the final form of P!=NP: formally proven, yet still inscrutable.

    These elements of Mathslop may still find their way into being used in Mathlib — much like the four color theorem, where I doubt many if any people people have hand verified each of the 1500-odd cases, yet it is now universally accepted as true and used as a basis for other work.

    Perhaps someday: “If I have seen further, it is because I have stood on the shoulders of silicon giants.”

  14. Comment on Do we want to stop all crime? in ~society

    unkz
    Link Parent
    I've undergone a few significant shifts in my set of ethics, so I recognize that even what I believe right now is probably wrong in many ways and won't reflect what I think in even 5-10 years.

    I've undergone a few significant shifts in my set of ethics, so I recognize that even what I believe right now is probably wrong in many ways and won't reflect what I think in even 5-10 years.

    10 votes

  15. Comment on How to prevent mold growth under weight mats in ~life.home_improvement

    unkz
    Link
    Those look like they are almost definitely from spilled liquid. I would say, if you spill liquid immediately pull up the mats and wash/dry them, and you won’t have an issue. And maybe clean once a...

    Those look like they are almost definitely from spilled liquid. I would say, if you spill liquid immediately pull up the mats and wash/dry them, and you won’t have an issue. And maybe clean once a year — in the gym I train at we have 20 people in there soaking the mats in sweat like a slip and slide, and it’s fine. We do spray and mop the top layer after every class, and pull it all apart annually to be sure.

    4 votes

  16. Comment on What are people's experiences with using Kagi? in ~tech

    unkz
    Link
    I used to use Kagi but kind of drifted away from it. Something interesting happened though which startled me: after I stopped using it, they started emailing me saying that they noticed I stopped...

    I used to use Kagi but kind of drifted away from it. Something interesting happened though which startled me: after I stopped using it, they started emailing me saying that they noticed I stopped using it, and so they were refunding my monthly subscription fee and that they hoped I’d come back some day. So, A+ for pro-consumer behaviour.

    88 votes

  17. Comment on Is AI profitable yet? in ~tech

    unkz
    Link Parent
    I think two factors Anthropic has almost zero mindshare along casual users OpenAI is the backend of Microsoft Copilot and Github, which is everywhere

    I think two factors

    • Anthropic has almost zero mindshare along casual users
    • OpenAI is the backend of Microsoft Copilot and Github, which is everywhere
    4 votes

  18. Comment on Outsourcing plus local AI will soon become more economical vs frontier labs in ~tech

    unkz
    Link Parent
    https://developers.openai.com/cookbook/examples/reproducible_outputs_with_the_seed_parameter With the caveat that this is only mostly reproducible due to fundamental issues with parallelism. It’s...

    I recall somebody being surprised they haven't allow seeding the random number generator. This is by design, because it would make the results deterministic and you could test different models apples to apples.

    https://developers.openai.com/cookbook/examples/reproducible_outputs_with_the_seed_parameter

    With the caveat that this is only mostly reproducible due to fundamental issues with parallelism. It’s a best effort but on relatively short responses you won’t find too much drift.

    2 votes

  19. Comment on I think Anthropic and OpenAI have found product-market fit in ~tech

    unkz
    Link Parent
    People used to be able to make an argument that compilers were less efficient than hand tuned machine code, but nowadays, I’m pretty confident that all in all but the most unusual circumstances an...

    People used to be able to make an argument that compilers were less efficient than hand tuned machine code, but nowadays, I’m pretty confident that all in all but the most unusual circumstances an optimizing compiler will far outperform a human.

    What this means I think is that our best efforts from our brightest minds in the machine language space are now spent on optimizing compilers instead of writing bespoke machine code to solve individual problems.

    This is what I see in the future: we will still have high level programming experts and mathematicians who know C or Rust or high dimensional mathematics in a deep way, but their efforts will be largely focused on improving coding agents who then in turn interface with “regular” developers, like compilers or interpreted languages do today.

    5 votes

  20. Comment on I think Anthropic and OpenAI have found product-market fit in ~tech

    unkz
    Link Parent
    This is an interesting problem for sure, and I’ve spent a lot of time thinking about it. On the other hand, we used to say that about assembly language programmers.

    This is an interesting problem for sure, and I’ve spent a lot of time thinking about it. On the other hand, we used to say that about assembly language programmers.

    4 votes