Prioritize YouTube questions
This page lets you prioritize questions from YouTube in the question queue so that when someone asks for a question on Discord it's a reasonably good one.
Hover over review levels to get a description of each. If you are unsure about what to categorize questions as drop by #general and ask for a second opinion. Once a page has been given a reviewed status it will no longer appear here.
Prioritize YouTube questions
2 questions prioritized and ready to be asked, 1792 questions which could be sorted! 1036 asked on Discord, 258 answered.
Highly upvoted YouTube questions
Why does the operational environment metric need to be the same one as the learning environment? Why not supervise cleaning 100% of the time during learning, then do daily checks during testing, then daily checks once operational. Expensive initially but the the 'product' can be cloned and sent out to operational environments en-mass. Motezuma (sp?) training with some supervisor (need not be human) in the training phase. Rings of training my children to put their own clothes on in the morning. No success so far.
Could you please make some video about generic intelligence concepts and AGI self-learning
Hey, love the vids.
I'm curious: could you in any future videos, recommend any recent literature regarding AI that is worth a read? I am sure you are familiar with "Superintelligence: Paths, Dangers, Strategies" by Nick Bostrom. Could you maybe give an overview of the ideas presented there? How accurate do you think it is? Are there any aspects of if that have become false in the most recent years with regards to the enormous progress in the AI field?
One of the most big shocks for me was that even thou N. Bostrom's ideas came to as a bit far fetched most of the times, even he underestimated some of the AI advancements. Like predicting AI Go victories with professional players to be at least 10 years away. Or that Poker playing AI will have a hard time beating our best. They already are...
Is there currently any research done in terms of biological AI? Gene sequencing influencing IQ or any such magic?
5:10
Rob, could you make a video about those philosophical problems? (I get this is not your area, but just a quick video enumerating them, for example)
What about adding a "contentedness" function? "My stamp collection is complete. Now I can rest."?
I want "later" (as in "more on that later...") to be "now". How long will I have to wait?
The one thing that I can figure out about AGI, is how can you make an AGI without letting it able to change its reward function to it is maximum happiness by just sitting around.
I have a question re: general intelligence AI not wanting to be corrected or upgraded. Would that correlate at all to human general intelligence? I'm thinking about it in the sense that as a child, I did not enjoy going to school and did not understand the value of it, while now as an adult, I enjoy learning new things. Would AI be able to reach a point of 'maturity' where it would perceive value in correction, or is that not a likely outcome?
Super interesting! If this kind of reward hacking exists in current AI, does that have any kind of serious implications if someone wanted to deploy one for the stock market, for example? Like would the AI seek to "cheat" and commit fraud or some gain insider info rather than play the stock market fairly?
Will the next video be called "Reward Hacking Revelations"?
This turns out to be an exploration of the human mind, of thinking processes and language usage. I think there shall be at some point a universally accurate language that expresses exactly what we mean, not what we say. Oh, wait, wouldn't that be math? So why not specify exactly the intended position of the red lego brick in relation to the black one? Just one example...
I could listen to you for hours. Also, where can we hear the full cover of Everybody Wants to Rule the World?
So the idea is that the mesa optimizer is smart enough to understand that it's being trained and will know when it is deployed? In the toy example, how would the mesa optimizer learn that it is safe to eat apples #3 to #5?
can you build an ai to edit the videos for you?
So who's working on an AI that operates in a user-access shell environment and gets rewarded for gaining root access?
Recent unreviewed YouTube questions
What happens when someone trains an AI variating the coin placement?
General question : can't we have a objective to predict what the user wants the objective to be ? That should atleast learn that we user probably don't want so just go topright but more like just collect that yellow thing as the former looks too dumb
Learning the wrong goal and continuing to persue it thinking it is the right goal? Sounds like the models need therapy.
Hang on... isn't the extremely strong "positiveness" on the coin, in the learning cases, simply a consequence of overfitting? It sees the coin *and* it's at the right-most corner. And so if this overfitted model sees that the rightmost wall is not to the right of the coin, then it has no reason to believe that the coin exactly is the thing that is the greatest thing in the universe. By this hypothesis, the rightmost wall of the map should (I think) also have a severely reduced positiveness score in deployment, because the wall is alone... without the coin... And it appears like that's the case? In addition, I bet if the randomly placed coin happens to be spawned at the rightmost corner of the map, then I'm sure for that particular run, the positiveness score will skyrocket for that particular run, even in deployment...
The best thing in the universe is the combination of a coin and the rightmost wall.
"Remember that distinction, from earlier videos? *shows footage from the 70s*"
