<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>llm on James Alvarez</title><link>https://jamesalvarez.co.uk/tags/llm/</link><description>Recent content in llm on James Alvarez</description><generator>Hugo</generator><language>en-gb</language><lastBuildDate>Mon, 09 Mar 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://jamesalvarez.co.uk/tags/llm/index.xml" rel="self" type="application/rss+xml"/><item><title>YouTubeToLingQ - turning YouTube videos into A2 German reading practice</title><link>https://jamesalvarez.co.uk/blog/youtube-to-lingq/</link><pubDate>Mon, 09 Mar 2026 00:00:00 +0000</pubDate><guid>https://jamesalvarez.co.uk/blog/youtube-to-lingq/</guid><description>&lt;p&gt;I've been stepping up with my learning German and currently one of my favourite tools is &lt;a href="https://www.lingq.com"&gt;LingQ&lt;/a&gt;. You can read upload text in your target language, and get an instant translation of specific words in the reader, and it keeps track of your vocabulary over time.&lt;/p&gt;
&lt;p&gt;LingQ has a YouTube import feature, but it's not reliable and you can't pick English videos to get them in German. I wanted a way to take any YouTube video, get the transcript, and have it translated into A2-level German, which is the level I'm at.&lt;/p&gt;</description></item><item><title>How useful are LLMs for historical research?</title><link>https://jamesalvarez.co.uk/blog/llms-history-research/</link><pubDate>Thu, 19 Dec 2024 00:00:00 +0000</pubDate><guid>https://jamesalvarez.co.uk/blog/llms-history-research/</guid><description>&lt;p&gt;I recently attended a Symposium at Sussex University titled 'AI and the Production of Historical Knowledge'&lt;/p&gt;
&lt;p&gt;&lt;a href="https://arxiv.org/abs/2311.08398"&gt;https://arxiv.org/abs/2311.08398&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://openreview.net/forum?id=xlKeMuyoZ5#discussion"&gt;https://openreview.net/forum?id=xlKeMuyoZ5#discussion&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://arxiv.org/abs/2411.09978"&gt;https://arxiv.org/abs/2411.09978&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://sociologica.unibo.it/article/view/19576"&gt;https://sociologica.unibo.it/article/view/19576&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://www.sciencedirect.com/science/article/pii/S1045235424000212"&gt;https://www.sciencedirect.com/science/article/pii/S1045235424000212&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;I ended up doing a bit of research into the very practical question of what do LLMs do well and not so well for researching history.&lt;/p&gt;
&lt;h4 id="the-bad"&gt;The bad&lt;/h4&gt;
&lt;h3 id="1-temporal-collapse"&gt;1. Temporal Collapse&lt;/h3&gt;
&lt;p&gt;LLMs often exhibit &amp;quot;temporal collapse&amp;quot; where they blend information from different time periods without clear demarcation. For example when describing a historical figure's views on technology, may inadvertently incorporate modern technological concepts. They may fail to recognize when certain concepts or terms didn't exist in the time period being discussed and so they may generate plausible but historically impossible connections&lt;/p&gt;</description></item></channel></rss>