I skimmed through the guide for instructors, and as someone with a M.Sc in NLP and machine learning, I think this course looks pretty impeccably designed. They include everything that I...
I skimmed through the guide for instructors, and as someone with a M.Sc in NLP and machine learning, I think this course looks pretty impeccably designed. They include everything that I immediately thought of as well as some stuff that it wouldn't have even occurred to me to include in a course like this. (Shoutout to speech act theory making an appearance!)
Two University of Washington professors (data science and biology) have developed an 18-lesson course specifically designed to help students navigate the world of Large Language Models. Unlike...
Two University of Washington professors (data science and biology) have developed an 18-lesson course specifically designed to help students navigate the world of Large Language Models. Unlike typical AI courses, this is approached from a humanities perspective and requires no technical background.
Each lesson is 5-10 minutes long and explores a core principle about LLMs, taking a balanced view that acknowledges both their impressive capabilities and serious concerns. The course includes instructor materials for classroom use, though it's also suitable for self-study.
This course makes a number of claims about LLMs, mostly substantiated by repeating the claim that "they just predict the next token" which hides a lot of complexity and controversy behind the word...
This course makes a number of claims about LLMs, mostly substantiated by repeating the claim that "they just predict the next token" which hides a lot of complexity and controversy behind the word "just". The course also takes a clear stance that LLMs cannot truly reason, understand, or have consciousness when all of these points are hotly debated in the field.
While I don't necessarily think it's bad to recommend sources with different opinions on AI, allowing people to form their own opinions on the topic by reading a breadth of different perspectives,...
While I don't necessarily think it's bad to recommend sources with different opinions on AI, allowing people to form their own opinions on the topic by reading a breadth of different perspectives, I think it's pretty absurd to claim Karpathy is a less biased source here. He created/owns his own "AI+Education" company and was a founding member of Open AI. He's got tons of incentives (financial and otherwise) to be biased towards bullish sentiment towards generative AI. I haven't engaged with his content before, so it's very possible that he still portrays a balanced and accurate view of things, but I think it's dishonest to present him as some neutral expert when he very obviously has skin in the game that the creators of this course do not.
I think you're right to be skeptical, and I didn't mean to present Andrej Karpathy himself as an unbiased figure, however I believe that if you watch the video you'll find that it is much less...
I think you're right to be skeptical, and I didn't mean to present Andrej Karpathy himself as an unbiased figure, however I believe that if you watch the video you'll find that it is much less biased than the linked article.
What I read of the teacher's guide on the linked site did a good job reflecting academic consensus and had a good balance of teaching the potential and limitations of generative AI. I think much...
What I read of the teacher's guide on the linked site did a good job reflecting academic consensus and had a good balance of teaching the potential and limitations of generative AI. I think much of what you perceive as bias is simplification of some of the more complex concepts either for the target audience or to market the course more effectively to non-technical educators like English teachers.
I haven't had a chance to watch the Karpathy video yet (though I have added it to my watch later playlist) but based on the chapter titles it seems to be aimed at a different, more technical audience than the linked course.
I tend to agree with this take. Claiming an LLM is just a fancy text predictor is the same as claiming a fully self-driving car is the same as a horse carriage because they both have wheels. I...
I tend to agree with this take. Claiming an LLM is just a fancy text predictor is the same as claiming a fully self-driving car is the same as a horse carriage because they both have wheels.
I would also like to point to the recent 3blue1brown videos on YouTube (I think there are a total of 7 now, but the last three tackle the LLMs specifically) which go into the mathematics of how certain concepts or 'understandings' (e.g. the similar difference between man/woman, king/queen, nephew/niece) are represented in the model, and where and how facts are stored.
I also definitely think 3blue1brown videos do very good jobs explaining how things like this work technically/mathematically. Though I haven't seen the newer content from him that goes into LLMs...
I also definitely think 3blue1brown videos do very good jobs explaining how things like this work technically/mathematically. Though I haven't seen the newer content from him that goes into LLMs specifically, I've watched his videos on other language model components and many of his videos covering calculus and linear algebra topics, which have been consistently good. I think they're intended for a different audience than the linked course, but they're phenomenal for learning how we get from math to these seemingly inscrutable models.
I started going through the course, and I like it so far. Thanks for sharing. I've learned a few things that will be beneficial for some projects I'm working on. I believe I am early to the course...
I started going through the course, and I like it so far. Thanks for sharing. I've learned a few things that will be beneficial for some projects I'm working on.
I believe I am early to the course because there are a couple videos that say coming soon at the end. However they are not necessary for the course, just nice to have.
I skimmed through the guide for instructors, and as someone with a M.Sc in NLP and machine learning, I think this course looks pretty impeccably designed. They include everything that I immediately thought of as well as some stuff that it wouldn't have even occurred to me to include in a course like this. (Shoutout to speech act theory making an appearance!)
Two University of Washington professors (data science and biology) have developed an 18-lesson course specifically designed to help students navigate the world of Large Language Models. Unlike typical AI courses, this is approached from a humanities perspective and requires no technical background.
Each lesson is 5-10 minutes long and explores a core principle about LLMs, taking a balanced view that acknowledges both their impressive capabilities and serious concerns. The course includes instructor materials for classroom use, though it's also suitable for self-study.
This course makes a number of claims about LLMs, mostly substantiated by repeating the claim that "they just predict the next token" which hides a lot of complexity and controversy behind the word "just". The course also takes a clear stance that LLMs cannot truly reason, understand, or have consciousness when all of these points are hotly debated in the field.
If you're looking for an introduction to LLMs without such a heavy bias I would recommend this great video by Andrej Karpathy: https://www.youtube.com/watch?v=7xTGNNLPyMI.
While I don't necessarily think it's bad to recommend sources with different opinions on AI, allowing people to form their own opinions on the topic by reading a breadth of different perspectives, I think it's pretty absurd to claim Karpathy is a less biased source here. He created/owns his own "AI+Education" company and was a founding member of Open AI. He's got tons of incentives (financial and otherwise) to be biased towards bullish sentiment towards generative AI. I haven't engaged with his content before, so it's very possible that he still portrays a balanced and accurate view of things, but I think it's dishonest to present him as some neutral expert when he very obviously has skin in the game that the creators of this course do not.
I think you're right to be skeptical, and I didn't mean to present Andrej Karpathy himself as an unbiased figure, however I believe that if you watch the video you'll find that it is much less biased than the linked article.
What I read of the teacher's guide on the linked site did a good job reflecting academic consensus and had a good balance of teaching the potential and limitations of generative AI. I think much of what you perceive as bias is simplification of some of the more complex concepts either for the target audience or to market the course more effectively to non-technical educators like English teachers.
I haven't had a chance to watch the Karpathy video yet (though I have added it to my watch later playlist) but based on the chapter titles it seems to be aimed at a different, more technical audience than the linked course.
I tend to agree with this take. Claiming an LLM is just a fancy text predictor is the same as claiming a fully self-driving car is the same as a horse carriage because they both have wheels.
I would also like to point to the recent 3blue1brown videos on YouTube (I think there are a total of 7 now, but the last three tackle the LLMs specifically) which go into the mathematics of how certain concepts or 'understandings' (e.g. the similar difference between man/woman, king/queen, nephew/niece) are represented in the model, and where and how facts are stored.
I also definitely think 3blue1brown videos do very good jobs explaining how things like this work technically/mathematically. Though I haven't seen the newer content from him that goes into LLMs specifically, I've watched his videos on other language model components and many of his videos covering calculus and linear algebra topics, which have been consistently good. I think they're intended for a different audience than the linked course, but they're phenomenal for learning how we get from math to these seemingly inscrutable models.
I started going through the course, and I like it so far. Thanks for sharing. I've learned a few things that will be beneficial for some projects I'm working on.
I believe I am early to the course because there are a couple videos that say coming soon at the end. However they are not necessary for the course, just nice to have.