Post B5NWrkXWMvs0GIru2i by jhlagado@jorts.horse
 (DIR) More posts by jhlagado@jorts.horse
 (DIR) Post #B5NWrkMt0TMZjJtPSy by jhlagado@jorts.horse
       0 likes, 0 repeats
       
       Redux: Open source is deadThe latest arguments are based on moral panic about AI security but they obscure the real reason: LLM generated software cannot be open sourced OR closed sourced because it cannot licensed at all.It can be either held as a trade secret or released as public domain. There is no in between so a lot of these discussions about licensing in the context of AI are frankly dishonest and disingenuous.
       
 (DIR) Post #B5NWrkXWMvs0GIru2i by jhlagado@jorts.horse
       0 likes, 0 repeats
       
       I realise that not all open source is being written by LLMs, far from it. What I am referring to are people and companies who are trying to obscure the nature of the software they are producing by falling back on licensing paradigms that no longer apply to them.
       
 (DIR) Post #B5NWrkotKLlT8Azm7M by ramin_hal9001@fe.disroot.org
       0 likes, 0 repeats
       
       @jhlagado  I have heard many large companies are requiring their employees to use LLMs for coding and other things.I assume this is because they have bought into the idea that:LLMs are a disruptive new technology comparable to the invention of the personal computer (I’ll include smart phones in the PC category)any company that isn’t using them will quickly learn that the technology they use is out-of-date and fails to keep pace with their competitors, andLLMs are a skill that anyone needs to learn, just as anyone needs to learn how to use a PC.Given the limitations of LLMs, and also the ill effects LLMs have on people that we are only just beginning to understand, I am highly dubious of all three assumptions.I think LLMs are less like the invention of the PC, and more like the discovery of radium. We are in the early stages of the discovery, when industry is struggling to find uses for it, not yet realizing in what ways it is harmful to people, and there is lots and lots of snake-oil products being sold by companies that used to have a reputation for producing safe and useful products.
       
 (DIR) Post #B5NWrlGBgrbCUplZfE by jhlagado@jorts.horse
       0 likes, 0 repeats
       
       @ramin_hal9001 The extremely rapid adoption of this technology by enterprises, despite its newness and unproven benefits and the near-total absence of skepticism, makes me think the main attraction of LLMs is literally their ill effects.Nobody is as excited about a technology that promises people liberation from drudgery and exhaustion. LLMs offer employers the promise of increasing the intensity of work and devaluing human labour power, two effects documented by Marx in Das Kapital Vol 1.
       
 (DIR) Post #B5NWrlboOStdZtsqMy by pfm@edolas.world
       0 likes, 0 repeats
       
       @jhlagado @ramin_hal9001 The sad thing is that benefits are proven, but we - the skeptical, politically-aware folks - may not like them.For managers, it's exactly as Ramin wrote: the disruptive technology that they can use as a slingshot for their career. Even if LLMs are known to generate code that's sometimes flawed, they hope for the slingshot to move them from their current positions to the next ones, so when the flaws of these decisions become obvious, they'll already be far from blast radius.Meanwhile for regular workers it's the promise of getting the job done quicker and avoiding tedious work like writing unit tests or boilerplate code. The people who buy this are of course the ones who don't know that there are better tech stacks than the one they're working with (e.g. requiring less boilerplate), or who consider worker unions a bad thing and therefore treat all opposition to what the employer decides a bad thing.The benefit for the first group is obvious. The benefit for the second group is: getting *some* results sooner than engineers who rely on their own brains (experience, skills, creativity, etc.).