LLMs Explained: Tokens, Embeddings, Transformers and More

Comments · 965

• SY

  @syntaxfm · 2 months ago · pinned

  <a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=114">01:54</a> - history of chatbots<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=261">04:21</a> - chat bot code<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=409">06:49</a> - black box thinking<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=461">07:41</a> - what are neural networks<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=513">08:33</a> - history of neural networks<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=627">10:27</a> - XOR neural network code<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=935">15:35</a> - what is tokenization<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=992">16:32</a> - history of tokenization<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1075">17:55</a> - tokenizer code<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1297">21:37</a> - what are embeddings<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1324">22:04</a> - history of embeddings<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1413">23:33</a> - what is a vector?<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1479">24:39</a> - embedding code<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1848">30:48</a> - history of transformers<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1953">32:33</a> - what is a transformer<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1991">33:11</a> - what is self attention?<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2065">34:25</a> - what is a feed forward network?<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2077">34:37</a> - what are stacked transformer blocks?<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2497">41:37</a> - what are softmax, temperature and top-p<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2574">42:54</a> - the autoregressive loop<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2608">43:28</a> - the context window<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2685">44:45</a> - the full picture<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2726">45:26</a> - what is fine-tuning<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2799">46:39</a> - RLHF<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2844">47:24</a> - tool calling<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2916">48:36</a> - the AI summit<br><a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=2991">49:51</a> - final thoughts

  118

• JA

  @JamesJosephFinn · 2 months ago

  This isn’t an episode, this is a full blown documentary!

  722

• JL

  @jlinkels · 2 months ago

  Really the most clear and thorough &nbsp;explanation on LLMs I have ever encountered.

  247

• AB

  @AbhinavKulshreshtha-v3q · 2 months ago

  Never knew that Claude is a homage to actual person.

  287

• AD

  @AdrianPilarczyk-cr6mu · 2 months ago

  When Syntax joined Sentry, I was skeptical. When CJ joined Syntax, I was skeptical. I&apos;ve obviously turned around on both of these a long time ago, but this episode made me realize how far Syntax has come. Really fantastic stuff, congrats. CJ is perfect for these technical deep dives.

  281

• SK

  @skepticusmaximus184 · 2 months ago

  @<a href="https://www.youtube.com/watch?v=YmLp8qe87A0&amp;t=1450">24:10</a> &quot;Happy and fridge are far apart.&quot; But only when there&apos;s no beer in the fridge. 😝

  56

• PA

  @paulcazza · 1 month ago

  This is the result of hours and hours of carefully planned out hard work. I am very grateful.

  24

• MY

  @mymusikrawks · 2 months ago

  As a professional programmer, but with little LLM knowledge, this was the PERFECT introduction to this for me. Loved the history and iterative method of moving towards the full idea of building these things from scratch. GREAT JOB!

  41

• RI

  @riplee406 · 3 days ago

  This is the best video I’ve seen that explains LLM from start to finish. The visualizations are top-notch and really make a difference.

  5

• FE

  @felixallistar · 2 months ago

  absolute banger. its so nice having everything tied together. <br><br>11/10 editing too.

  210

• NA

  @nachiketpatil2604 · 2 months ago

  It&apos;s amazing how detailed the explanation is despite being so concise.

  26

• SA

  @SamuelSani-t8s · 2 months ago

  This is mind blowing, you gave me history, statistics, programming, visualization, presentation. All in one and saying I understand how LLMs work now is an understatement to this video.

  43