从混乱到确定性——重塑叙事空间
2 分•作者: pyeri•3 个月前
今天早上喝着印度奶茶时，我意识到当前人工智能热潮很大程度上源于“人工智能叙事”。大型语言模型（LLM）技术本身就像许多之前的数字工具一样，但关于“人工智能是未来”、“学习它否则就灭亡”、“机器很快将取代人类”等等的喧嚣，让它持续成为新闻焦点，并无中生有地制造轰动。然而，人们却为此耗费精力、牺牲睡眠，这反而成了一个问题。许多本本分分的企业也因此被误导，认为这项技术能创造奇迹，而那些兜售这些“桥梁”的人，正在制造巨大的误导。
对于我们这些厌倦了这种陈词滥调的人来说，与其对抗它，不如尝试将叙事转向决定论。几个世纪以来，我们一直被训练去避免甚至恐惧宿命论（决定论的一个负面方面），并对未来寄予改善和繁荣的希望。有希望并没有错，但纵观有记载的历史，“未来将是辉煌的”这一主题似乎并不那么光明，至少对于许多与基层联系的人来说是这样。
为了对抗这种叙事，我们需要将决定论转变为一种积极的力量，或者至少是一种比替代方案更可取的选择。即使在大型语言模型领域，大多数实用的东西都出现在低端、开源和本地化的大型语言模型领域，这些领域的方法更具决定论色彩。我们能做些什么，来让人们重新对使用 C、Java 和 Python 进行常规的决定论编程产生兴趣呢？
查看原文
What I realized today while sipping my morning cup of chai is that a large part of the present AI hype cycle is just about the 'narrative of AI'. The LLM technology itself is just a digital tool like many others that came before it but all this chatter about 'AI is the future', 'learn it or perish', 'machines will replace humans soon', etc. keeps it in the news and creates a burger out of nothing. But folks lose their energy and sleep over this which becomes a problem. And many a benign enterprises are falsely led to believe that this technology can do wonders, it's a great disservice being done by those selling these bridges.<p>For folks like us who are fed up of this slop narrative - instead of fighting it, it's better to try and shift the narrative towards determinism. For centuries and millennia, we have been trained to avoid and even fear fatalism (a negative aspect of determinism), and look towards future with hopes for betterment and prosperity. There is nothing wrong with having hope but for the first time in known history, the 'future will be glorious' theme isn't looking so bright, at least to a large number of folks connected to grassroots.<p>In order to push back against this narrative, we need to turn determinism into a force of good, or at least a force preferable to the alternative. Even in the LLM space, most of the utilitarian things are happening in the low-end, open source and local LLMs niches which are more deterministic in approach. What can we do to bring back people's interest in regular deterministic programming with C, Java and Python?