We are scarily close to realtime personalization of video which if you agree with this NeurIPS paper [1] may lead to someone inadvertently creating “digital heroin”
> We further urge the machine learning community to act proactively by establishing robust design guidelines, collaborating with public health experts, and supporting targeted policy measures to ensure responsible and ethical deployment
We’ve seen this play out before, when social media first came to prominence. I’m too old and cynical to believe anything will happen. But I really don’t know what to do about it at a person level. Even if I refuse to engage in this content, and am able to identify it, and keep my family away from it…it feels like a critical mass of people in my community/city/country are going to be engaging with it. It feels hopeless.
I tend to think that it leads to censorship, and then censorship at a broader level in the name of protecting our kids. See with social networks where you now have to give your ID card to protect kids.
The best way in that case is education of the kids / people and automatically flag potentially harmful / disgusting content and let the owner of the device set-up the level of filtering he wants.
Like with LLMs they should be somewhat neutral in default mode but they should never refuse a request if user asks.
Otherwise the line between technology provider and content moderator is too blurry, and tomorrow SV people are going to abuse of that power (or be coerced by money or politics).
At a person / parent level, time limits (like you can do with web filtering device for TikTok), content policy would solve and taking time to spend with the kids as much as possible and to talk to them so they don’t become dumber and dumber due to short videos.
But totally opposed that it should be done on public policy level: “now you have right to watch pornography but only after you give ID to prove you are adult” (this is already the case in France for example)
It can quickly become: “now to watch / generate controversial content, you have to ID”
That doesn't work when the Chinese produce uncensored open weight models, or ones that can easily be adapted to create uncensored content.
Censorship for generative AI simply doesn't work the way we are used to, unless we make it illegal to posess a model that might generate illegal content, or that might have been trained on illegal data.
> Censorship for generative AI simply doesn't work the way we are used to, unless we make it illegal to posess a model that might generate illegal content, or that might have been trained on illegal data.
Censorship doesn't work for stuff that is currently illegal. See pirated movies.
It saddens me to think that the efforts so far hasn't been it. Maybe I should try my hand at "closing the loop" for image generation models.
Could it destroy the society? The humanity had lived through bunch of such actual substances, and always got bored of it in matters of decades... those risk talks feel a bit overblown to me.
Nice one! I've been doing something similar with GPT to help learn Spanish:
Whenever I type something in Spanish you translate what I think I meant into English, point out my mistakes if any, and also respond with the corrected sentence
I've been doing advanced calisthenics for a few years now and have mastered some relatively advanced skills such as the one-arm pull-up, touch front-lever, handstand push-ups, and straddle planche. (I'm also a licensed PT in the UK but use the recommendations here at your own risk)
For beginners, if you are thinking of skills-based training I'd recommend focussing on 1-2 skills at a time and developing them (e.g., front-lever and handstand push-ups). A cool thing about calisthenics is that even the most advanced exercises start with simple progressions. Also, contrary to popular belief, you can develop a lot of muscle mass through body-weight exercises.
There are many great instructors out there, some I'd recommend off the top of my head are Simonster Strength, Eryc Ortiz, Gabo Saturno, Coach Sommer, and Victor Kamenov. Simonster has some really good tutorials for beginners as well as advanced athletes which I've used (https://simonsterstrength.com/project-calisthenics/ plus loads of great workouts on his insta/youtube).
In terms of split, there's a few common approaches like Push/Pull/Leg or Upper/Lower and unless you are doing supersets I'd recommend 3-5 exercises for 3-5 sets each with 2-3 min rest between sets. The number of reps/time under tension will depend on the exercise, your strength level, and your goals.
To give an example, a really simple split for beginners could be something like:
- Leg Day: Warm up, Squats, Lunges, Hip Thrusts, Calf Raises, Warm down, Stretch
- Rest Day: Cardio or Light Stretching
Just go on youtube if you are unsure about any of the exercises. Also, feel free to use a band or look for easier progressions for any of the exercises and
And this is with human generated content. Wait until computer generated content can be used to optimize human engagement. I've thought about this a lot and it scares me. We may need more regulation.
LOL, I suppose "fiery email domain" could mean outlook/exchange if we think in terms of reliability. I updated it to "Microsoft's not cold email domain acquired in 1997".
Made https://anonfeedback.io last weekend after looking for a way to get anonymous feedback from colleagues and not wanting to sign up for any service.
[1] https://neurips.cc/virtual/2025/loc/san-diego/poster/121952