Both sides of the argument are marshalling LLMs to make their case.

争论的双方都带领着大型语言模型来证明他们的观点。

That's why grounding the LLM in data is so important.

这就是为什么以数据为基础的 LLM 如此重要。

Large Action Models, as some call such LLMs, could change that.

大型行动模型（有些人这样称呼法学硕士）可能会改变这一现状。

LLMs will also allow digital twins to adapt.

LLM 还将允许数字孪生进行适应。

And some are now seizing on the capacities of LLMs to attack Chomskyan theories anew.

其中一些人现在正利用大型语言模型的能力来对乔姆斯基的理论进行新一轮进攻。

An llm is of little use to someone picking asparagus.

LLM 对于采摘芦笋的人来说用处不大。

That leaves open the possibility that children have an inborn tendency to grammar, making them far more proficient than any LLM.

这就带来了一种可能性，那就是儿童天生就有语法倾向，这使他们比任何大型语言模型都更擅长语言。

Further distortions are possible when a pretrained LLM is " fine-tuned" .

当对预训练的 LLM 进行“微调”时，可能会出现进一步的扭曲。

The generator then acts like a normal LLM and answers.

然后，生成器就像正常的 LLM 一样运行并给出答案。

And indeed, LLMs seem to learn nested, hierarchical grammatical structures, even though they are exposed to only linear input, ie, strings of text.

事实上，虽然大型语言模型只接触了线性的输入内容，即文本字符串，但它们似乎学会了嵌套的层级语法结构。

Perhaps, then, the problem is that accuracy is too much to ask of LLMs alone.

那么，问题可能在于，单单对法学硕士 (LLM) 的要求准确性太高了。

One is that LLMs are not built to have perfect recall in the way a search engine or encyclopedia might.

一是 LLM 并不像搜索引擎或百科全书那样具有完美的回忆能力。

One of the wildcards is open-source AI, which has proliferated since March, when LLAMA, the LLM developed by Meta, leaked online.

通配符之一是开源 AI，自 3 月以来激增，当时由 Meta 开发的 LLAMA 在线泄露。

But other researchers have tried to train an LLM on a database of only child-directed language (that is, of transcripts of carers speaking to children).

但其他研究人员试图用只面向儿童的语言数据库(即照顾孩子的人对儿童所说话语的文字稿)来训练大型语言模型。

(laughs) - So, that takes us back to prompt engineering of noticing how to be a better prompt engineer of an LLM.

（笑）-所以，这让我们回到提示工程，注意如何成为一名更好的法学硕士提示工程师。

This reduces hallucinations by letting the LLM play to its strengths—summarising and paraphrasing rather than researching.

这可以减少幻觉，让法学硕士发挥其优势——总结和解释，而不是研究。

Even with the best algorithmic and architectural antipsychotics available, however, LLMs still hallucinate.

然而， 即使有了最好的算法和架构抗精神病药物，法学硕士仍然会产生幻觉。

Another way to detect hallucination is to train a second LLM to fact-check the first.

检测幻觉的另一种方法是训练第二个 LLM 对第一个进行事实核查。

Within six months of its launch, Microsoft released ChatGPT-powered Bing; Google demoed its latest large language model (LLM), Bard; and Meta released LLaMA.

推出后六个月内，微软发布了由 ChatGPT 支持的 Bing；谷歌演示了其最新的大型语言模型（LLM）Bard； Meta 发布了 LLaMA。

One clue that a hallucination is under way lies in how an LLM picks words.

判断幻觉是否正在发生的一个线索在于法学硕士如何选择词语。