(Workshop Series) #4: Autofeedback
Added 2023-09-03 18:30:02 +0000 UTC
Comments
After watching this video and the previous videos, I get the impression that you think this is the future of tvl. Coming from a computer science and machine learning background, I find this tricky. Machine learning may the big hype of the moment, but it is actually just one tool among others. As you point out, conventional theories have limitations: Maybe the other side fails to implement it. Be it because of the complexity or because the other side does not understand the intended meaning of your words. Without doubt, machine learning can do amazing things. ChatGPT could not exist without it, speech recognition would be feasible, but much more complicated and less accurate. I find it intriguing you adapt the concept to tvl. But machine learning also does have it's shortcomings: - It depends very much on your training data and your starting situation. Of course, you can do them at random, but then your success will also be random. It is much better to base it on a theory about the subject. For example, you will learn to ride a bike much faster if you know roughly how to do it from up front instead of searching through the entire chaotic spectrum for solutions. - It is very time-consuming and resource-intensive, as it is essentially a matter of trial and error. Depending on the initial setup, it can take a very long time or leave you right at the doorstep. - Machine learning often results in a black box (here: internal feelings) instead of understanding what the solution really is. Making it hard to optimize further unless you analyze it conventionally. - It can lead to you finding a solution that seems to work, but is not the right way: Yes, almost all children learn to speak. But some end up with speech therapists because they don't do it right or healthy. I was such a case. So i guess a teacher should not only inspire your targets but also check up your solution is doing it in a healthy way. The transmission of knowledge through complex language and text may not be infallible, but it is one of the main reasons why we humans are so much more successful than animals. We stand on the shoulders of giants. We don't have to research everything from scratch, but can build on it. That said: Yes, your conventional material didn't quite get me to my overall goal in the last few years. Often I felt like I couldn't fully grasp what you were trying to tell me. You are very right about the shortcomings of language in this area. But it was a valuable starting point and valuable inspiration, for example for this new concept here. I am glad to have had it. So I would invite you to consider ML not as the superior solution, but as another tool when the conventional solution doesn't work for an individual that seemingly does not share your mindset. In computer science, I would first try to think of a conventional approach before looking at ML. Even if it doesn't work, it puts me in a better ML starting position than if I didn't.
Laura
2024-06-30 21:36:41 +0000 UTCFor the first few videos of this series I was thinking "isn't it overkill to be analysing things in this much detail?" and this is the one that made everything finally click. I'd already started taking on board things like the importance of developing hearing and being guided by the sensory rather than physical side of things but this finally let me appreciate the basis for all of that and properly commit to it. I've been doing voice stuff for four years or so now, including teaching it, and I suddenly feel a lot more clarity about how I should practice and how I should teach.
Ruskin
2023-09-30 13:48:53 +0000 UTC
![SexFM [DO NOT DONATE]](https://xaiju.com/istorage/71004.jpg)
