ByteCode.News
Submit RSS Atom Sign in

Back to the channels

#java

  1. * mindCrime joined #java
  2. * MonsterAbyss joined #java
  3. * X-Scale joined #java
  4. * jjj333_p joined #java
  5. * Fiji_ joined #java
  6. * jiffy__ joined #java
  7. * five618480339176 joined #java
  8. * hwpplayer1 joined #java
  9. * MonsterAbyss joined #java
  10. * ferdna joined #java
  11. jreicher dreamreal: very late response, but Ekrem didn't use the word "tacit"; Polanyi did.
  12. * stfstfm joined #java
  13. * Inline joined #java
  14. * domicron joined #java
  15. * lostlazy_ joined #java
  16. * Nnavd joined #java
  17. * stfstfm_ joined #java
  18. * stfstfm joined #java
  19. * stfstfm joined #java
  20. * stfstfm_ joined #java
  21. Swayze oh we discussing that lol heres my conclusion: You can read and memorize 100 books on how to play the piano but that does not make one a master piano player
  22. Swayze the end
  23. Para Turns out playing an instrument is hard if you're tone deaf.
  24. jreicher Para: I assumed that made it very easy
  25. Swayze When doing interviews for junior software devs I found much more value in having people "read code" back to you and explain what it does (in an interview situation this has quite a few benefits and I recommend it as an approach, see https://freakingrectangle.wordpress.com/2022/04/15/how-to-freaking-hire-great-developers/). This is the exact skill that's at risk when considering the cognitive offloading vs OFF
  26. javabot Swayze's title: "How to Freaking Find Great Developers By Having Them Read Code | Freaking Rectangle"
  27. Swayze surrender discussion. The danger of AI we all feel instinctively is that surrender is too easy, and the only checks and balance we have in place is trusting that people are being responsible and not surrendering to AI. Perhaps one avenue is that since writing the code is potentially quicker, we should spend a lot more time in PR reviews and discussing the code we push measuring its fit for purpose, qualityOFF
  28. Swayze and if its a concern, to discourage surrender.
  29. jreicher FWIW my feeling these days is that the desire to "surrender", as you put it, is not new, because the problem is not new. AI slop might be new, but slop isn't. Losing expertise to AI might be new, but losing expertise isn't. The problem is the same as it's always been: where are the quality controls?
  30. jreicher The desire to surrender is, I think, just the fatigue from having to maintain quality in an environment where your colleagues might not feel the same, or might not even understand what quality is.
  31. Swayze thats one possible path to surrender
  32. Swayze addressing all of them is the challenge
  33. Zapek showing https://openjdk.org/legal/ai to people pushing too much AI slop had some chilling effect
  34. javabot Zapek's title: "OpenJDK Interim Policy on Generative AI"
  35. nevet OpenJDK Interim Policy on Generative AI
  36. jreicher I find that quite interesting
  37. * shsprk joined #java
  38. * yeahitsme joined #java
  39. * shsprk joined #java
  40. * five618480339176 joined #java
  41. * MikeBux joined #java
  42. * canton7 joined #java
  43. * X-Scale joined #java
  44. dreamreal jreicher: true BUT I'd say that Ekrem's use of polanyi sort of transitively migrates the "use of the word"
  45. dreamreal Zapek: yeah, but it's nothing new
  46. dreamreal I mean, the desire to surrender is .. complicated
  47. dreamreal Change is scary... and who owns the code generated by a compiler? We run through tools *all the time*. Provenance is complicated; assembly is "your code" assuming you wrote it by hand, machine code certainly is by the same token (it's not even an assembler that manages it), but...
  48. dreamreal AI is just another tool, another layer, with admittedly unclear provance and guard rails that are far more illusory than one would hope still
  49. * MikeBux joined #java
  50. * yano joined #java
  51. Zapek and our previous tools didn't require a "temperature" variable somewhere and didn't give a different output each time they were run with the same input
  52. * GreenResponse joined #java
  53. jreicher dreamreal: I don't think it's about provenance; I think it's about accountability/responsibility. Rightly or wrongly there's a perception that the way people use AI is not the same as the way they use other tools, and that there is a kind of abdication of responsibility that they don't engage in with non-AI tooling. (Note I am emphasising it's about the people and the use, and not about the tool)
  54. * ForeverDreaming joined #java
  55. dreamreal I don't think that perception is wrong at all.
  56. Inline hmmmm
  57. Inline IP and AI on active litigation
  58. dreamreal It's defintely a thing; I'm expecting a check from it myself.
  59. Inline don't use AI for coding
  60. Inline so you have a reserve of coders who can code without AI and you use that, when you run out of it you run out of it, means code that is working right now will work forever that way, but there won't be any progress outside that scarce resource left, and once it's over there'll be only a new era left with AI-code only and no IP
  61. dreamreal I think that's a stupid conclusion, but defensible.
  62. Inline hmmm, why stupid ?
  63. dreamreal Because it's saying "there's a class of tools and I will not use it." It's the Ludd position.
  64. Inline well that's what the above interim policy tells you
  65. dreamreal Sure. And I think that's ... difficult to defend.
  66. dreamreal There are situations where the logic works: like, using java for flight control is prohibited, right?
  67. dreamreal Do you know why?
  68. Inline because of exceptions and interrupts ?
  69. dreamreal no. Because what runs isn't what's delivered.
  70. Inline what do you mean ?
  71. Inline oh, you mean the whole jvm has to be there too ?
  72. dreamreal Bytecode rarely runs in a JVM. What *runs* is morphing JIT code, based on the happy path. That's difficult to test in an absolute sense.
  73. Inline hmm, ok
  74. dreamreal non-JIT code might be permissible, but it's slow enough that you wouldn't bother.
  75. dreamreal So: JIT is bad, right? I mean, it's a tool that is difficult to verify!
  76. dreamreal It's too hard to understand what the JIT *will* do for mere mortals; Josh Bloch and a few others can do it, but *I* can't, and apart from those few guys, I know nobody who can. I'm sure others exist, but they're rare.
  77. Inline so you say, we have other similar problems long before AI
  78. * Ragnor joined #java
  79. * Betal joined #java
  80. dreamreal Yes, AI is just another tool. It's a dangerous one - it amplifies errors like this and requires a lot of care - but that's... the scope of the problem. It's a tool. A big one with a lot of power. C'est la vie. Use with caution, just like every other tool, and use with the caution it deserves, which is *greater* than the caution for a compiler, for example.
  81. Inline hmmm, ok maybe it resolves as: AI is no IP, but any IP code which is augmented by non IP code is counted as IP still....
  82. Inline no idea
  83. Inline that would be one way i think
  84. Inline or it's the other way around as they frame it....
  85. dreamreal The legal issues are unresolved right now. I'm not sure #java is going to be able to resolve them, even if it was the right community to do it, and it's not.
  86. dreamreal I just don't think it's a candidate for "oh no, it's hard, throw it out"
  87. dreamreal That's the Luddite position, and while I understand Ludd, I think the Luddites were wrong. Progress is change, and change is scary, and sometimes change is wrong - there're a lot of horrifying cases where "let's fix this" had a death toll in the millions - but change in itself is something to EVALUATE.
  88. dreamreal The Luddites ended up losing an entire generation of expertise, sure... but we no longer can detect ship artisans by counting how many toes they've lost to adzes, and yes, I knowl the Luddites were textiles, not shipbuilding, but shipbuilding had the same mechanic, and the Luddites had their own environmental victims, they're just less dramatic. :D
  89. Inline but this kills the other side, when non IP AI code is worthless, everyone should be supposed to start from IP and that means again that you have done something yourself upfront which counts as IP
  90. Para fwiw AI cannot produce copyrighted works within EU legislation. There's upcoming more targeted legislation coming, but it'll still take a few years + probably isn't doing anything to this particular issue, as that is not consumer friendly.
  91. Inline eheh
  92. Inline so a small part IP + huge part non IP would still count as IP
  93. Inline yah, mixed code can be interpreted anyway, and it's up to the legislator of the country to decide on that, so it's not something which is to expected to be handled uniformly around the world
  94. Inline especiall: who cares if IP dies, and i still can make money with non-IP
  95. Inline trends trends..... we don't know how it will evolve....
  96. dreamreal I don't think non-IP AI code is worthless. IP is IP. An AI can be used to work on code, and the IP aspects are not the AI's domain.
  97. * magla joined #java
  98. Inline and since it would be non-IP AI code, you wouldn't be able to sell it directly, only as a service, i.e. if you service it right, nobody will ask what the code is/was
  99. Inline it's an internal bookkeeping thing of yourself so to say
  100. Inline it's worthy only as it brings the service otherwise on it's own has no value
  101. Inline welp, the prospects of the future are a little different it seems
  102. Inline if nobody is willing to pay for services or can't afford it....
  103. Inline bah, i get headaches
  104. dreamreal The path to the future is a maze of twisty passages, all alike... and we, as humans and as humanity, have a tendency to seek out tribal alliances as bets against solutions, and we overcommit to those tribes.
  105. * shsprk joined #java
  106. * stfstfm joined #java
  107. * stfstfm joined #java
  108. * stfstfm_ joined #java
  109. * X-Scale joined #java
  110. * rvalue- joined #java
  111. * stfstfm joined #java
  112. * stfstfm joined #java
  113. * troydm joined #java
  114. * fstd joined #java
  115. * shsprk joined #java
  116. * iwtga joined #java
  117. * stfstfm_ joined #java
  118. * tomaw_ joined #java
  119. * kusanagi joined #java
  120. * stfstfm joined #java
  121. * jiffy__ joined #java
  122. * gurrkin joined #java
  123. * SJrX joined #java
  124. * mindCrime joined #java
  125. * X-Scale joined #java
  126. * hwpplayer1 joined #java
  127. * stfstfm_ joined #java
  128. * fgarcia joined #java
  129. * geli joined #java
  130. * llanhmock joined #java
  131. * jreicher joined #java
  132. * jiffy__ joined #java
  133. * domicron joined #java
  134. * Exa joined #java