pull down to refresh
35 sats \ 1 reply \ @BlackDog 18 Dec \ on: YouTube Allowing Creators to Opt-In To AI Training on Their Content tech
From the AI perspective.. Many AI research centers (big tech, small tech, universities, ..) are downloading YT for training data. There are plenty of public datasets based on the YT. Especially if you are a celebrity, you are in such dataset.
There is push for model training data transparency (especially from over regulated EU). Meaning any model should clearly state, what was the training data (meaning all particular videos).
Next GDP (personal data protection) comes to play. You as the person in the content may share some personal data in your videos. It is not only your credit card number, address, phone, political preferences ... :) but it is also your face and voice (biometric information). And theoretically you may ask the author of the model to remove you from the training data (this is your GDP right) -- if the model is transparent you know it. In ideal world, they should do it. GDP is a bit complex, so there are ways you lose the right to so, but in very specific conditions.
Finally it also depends on the type of the model. If it is classification model (speech recognition, object detection in videos), you (as training data sample) are on safe side. It will just make subtitles of your voice bit more precise (you may want). But what about model for biometry ID? If you became a person of interest for someone, model pretrained on tones of your data from YT will work a bit better (they will need your data to train the final detector of your person anyway, but why make it easier for them giving your data also for pretraining). Finally, the generative models. Do you want AI to generate faces and voices similar to yours?
Regarding the opt-it now. I expect you are giving your consent to Google. So they can comply with EC regulations (I'm not sure how it works in US). They may sell the data to others (I'm not sure if it is align with YT license) -- I mean they sell the transparent data which has value! I'm not sure how this transfers into open data and open models.
There may be a regulation "war" in future. EC can ban US models due to non-transparency and eventually fine big tech for it. We saw/see EC vs MS, Apple, ... (as EU cannot win the tech race due to stupid regulations, they will profit using fines on noncompliance with the stupid regulations),
I guess that in future, the opt-in will expand into types of models you agree to be trained on you and also to whom you allow to do so (Google, big tech, for profit, non profit, ...). And also they will start to pay you. Ideally, when I train a model, I'll pay a fee to all the "training data".
So.. not to opt-in now will make pressure to be paid in the future. As the data is oil of the todays. Why to give it to Google for free and let them earn even more on it now.
(Sorry for longer post. I'm in the field and this is complex topic).
Few days back, I re-setup by full node. I went form RPi 4 + Umbrel to ASUS NUC (32GB, 2TB). I tried Start9 but went back to Umbrel. My concern is, that Start9 is a bit dead (looking into Git statistics). I want something with development. So far the node works pretty well. The RPi was sometime lagging.
Excellent question. Me as father of two young teens (and working in IT and ML and sitting in EU) is thinking a lot about this. My conclusions are:
- learn English (and may be Spanish). OK, we have all of this AI translation now, but you should be able to communicate around the world.
- learn skills (job) you can use world wide. IT is among the best, can be done remotely but yes, there is high competition (double edge sword). Please not select some really local job (as weird mother tongue language dialect linguists).
- be the best in your skills, which implies to learn new things all the time (to be on the edge).
- it is good to combine 2 areas .. this makes you quite unique.
- critical thinking, reasoning = math, logic. on the other hand some creativity is also needed.
- physical job is very OK. (if the above is met)
- understand health and money (keep you healthy and "rich"), do not let others and state to screw your life.
Why? Because the world is changing faster and faster. The time where you was OK with stuff from you school (even university) for your entire life is history. If you want to be competitive, you need to learn. I see it on myself in IT. I'm doomed by youngsters.
Why? Because I do not think EU will be good and safe place for their entire life. These times passed. So they should be ready to relocate = have compatible skills.
Now the most difficult part is how to implement this. But I cannot force them, so I'm explaining it and reasoning and doing gradually smaller and smaller decisions in their life. I'm trying to identify their strong abilities and emphasize them. Also randomly challenge them with new things so they can explore some new skills they like.
Am I doing it right? No. :) I just hope they will not screw their life by a single very bad decision totally :).
Yes, 44 now. I went through quite significant declination of health 3 years back. Till then, I was feeling quite good (not much sport, reasonable food), but suddenly, bang. My body told me "Hey man, you are not young anymore". So I'm learning how to live with it. Small morning exercises (yoga like), and some calinestatic. Listening to podcast about how to improve life (or limit really harming stuff like junk food, carbs, etc.).
I'm also fighting with brain fog. I'm not able to focus 8 hours/day 5day/week (IT job). I admitted, that younger kinds are smarter, faster than me. It was a bit painful, but... yes, that is the life. Now, I switch from mode "I'm can do and be what ever I want." to "I'm happy that I can catch the modern IT train and I can be still useful for others". It sounds depressive, but I need to be picky there I invest my energy as the resources are way more limited.
BTW there was some study recently, showing that aging has 2 steps (most noticable). After 40 and after 60 :)
GENESIS