New ask Hacker News story: Could Google Bard Disrupt Universities?
Could Google Bard Disrupt Universities?
2 by daly | 2 comments on Hacker News.
Step 1 Google collects a bunch of playlists. Bard classifies them as "networking", split by sub-topic (queueing theory, congestion handling, DNS, TCP/UDP, hardware types, wireless vs wired vs quantum, etc.) Step 2 is for Bard to summarize each lecture and create a 1-page summary of the important points as well as a syllabus of each course. Step 3 is a use the lectures as reinforcement learning from human feedback (RLHF) so Bard is able to interactively answer questions about networking before, during, and/or after the lectures. Step 4 is for Bard to create (or find) github repositories related to the course as well as the whole subject of networking. So now Bard IS the teacher. Iterate by topic. Step 5 is to have Bard create a "canonical version" of network lectures that it has self-generated, curated by experts in the field. Now you've disrupted all the Universities. I suspect that a small team could have most of this working within the next year or two using Bard doing voice recognition in lectures as well as Bard generated voice, video-from-text to illustrate the lectures, and even generate and correct homework problems unique to a particular user's weak areas of understanding based on interactions. Bard could be adaptive enough so that when I don't remember Little's Law it can dynamically insert a quick tutorial at the point in question. Companies could use such courses instead of resumes or interviews. You need to show a passing grade for a course in their problem area. People could easily transition from their current job to another job by taking Bard courses that target the employer. Everybody wins. Having deep control of youtube puts Google ahead of OpenAI. Skate where the puck is going, not where it has been. -- Wayne Gretzky
2 by daly | 2 comments on Hacker News.
Step 1 Google collects a bunch of playlists. Bard classifies them as "networking", split by sub-topic (queueing theory, congestion handling, DNS, TCP/UDP, hardware types, wireless vs wired vs quantum, etc.) Step 2 is for Bard to summarize each lecture and create a 1-page summary of the important points as well as a syllabus of each course. Step 3 is a use the lectures as reinforcement learning from human feedback (RLHF) so Bard is able to interactively answer questions about networking before, during, and/or after the lectures. Step 4 is for Bard to create (or find) github repositories related to the course as well as the whole subject of networking. So now Bard IS the teacher. Iterate by topic. Step 5 is to have Bard create a "canonical version" of network lectures that it has self-generated, curated by experts in the field. Now you've disrupted all the Universities. I suspect that a small team could have most of this working within the next year or two using Bard doing voice recognition in lectures as well as Bard generated voice, video-from-text to illustrate the lectures, and even generate and correct homework problems unique to a particular user's weak areas of understanding based on interactions. Bard could be adaptive enough so that when I don't remember Little's Law it can dynamically insert a quick tutorial at the point in question. Companies could use such courses instead of resumes or interviews. You need to show a passing grade for a course in their problem area. People could easily transition from their current job to another job by taking Bard courses that target the employer. Everybody wins. Having deep control of youtube puts Google ahead of OpenAI. Skate where the puck is going, not where it has been. -- Wayne Gretzky
Comments
Post a Comment