1,191 karma · joined February 8, 2014
edit: same thing as mentioned by ArtWomb in another comment >" This would make a cool art installation. If you can get access to one of those giant screens that you see demo'd at E3. You can visualize the entire Land of Hyrule from Lengend of Zelda. With a tiny Link navigating his way through the world ;) "
There has been recent focus on ML research trying to win or beat games. What if the goal was merely to have an agent that would manage to explore the largest game "area" possible? Has that been done before?
- Surviving Mars (Build and manage a colony on mars)
- Kerbal Space Program (Build spacecraft and send them into space)
- Gratuitous Space Battles (Place units and give them commands in space battles)
"Question: What is the answer to life, the universe and everything?
Aristo's Answer: 42
Confidence: 57.03%"
See excerpt here: https://twitter.com/nntaleb/status/892370148476190720
I recently read Sapiens by Yuval Noah Harari, where he thoroughly debunks this notion that a transition from a hunter-gatherer to an agricultural life-style brought more leisure time. I found it compelling, but don't have the arguments at hand at the moment.
On 1v1 saying that the winner is the one who concedes the least or that scores the most is equivalent.
Whereas on a 3 sided game it is not equivalent. See this example below (if you change the rule a different team will have won the game):
Team X: Scored/Conceded
Team A: 4/5
Team B: 5/3 (winner if highest scorer wins)
Team C: 1/2 (winner if lowest "conceder" wins)
[Edited to make the total scored and conceded add up to the same number.]http://aaa.t0.or.at/documents/aaarules.htm
> "Meanwhile the penetration of the defence by two opposing teams imposes upon the defence the task of counterbalancing their disadvantage through sowing the seeds of discord in an alliance which can only be temporary. This will be achieved through exhortation, body language, and an ability to manoeuvre the ball and players into such a position that one opposing team will realise that its interests are better served by breaking off the attack and allying themselves with the defending team. Bearing in mind that such a decision will not necessarily be immediate, a team may well find itself split between two alliances. Such a situation opens them up to the possibility of their enemies uniting, making maximum use of this confusion. 3-sided football is a game of skill, persuasion and psychogeography."
>"We managed to get about 9 lathes out of each 2x4 on average (sometimes 10, sometimes only 8). Remember, in planning for this, that the saw blade has thickness, and that all the sawdust has to come from somewhere."
Rewards for positive actions have a few side-effects that I have observed though:
1) Individuals start expecting positive reinforcement for everything they do. Doing it for the external reward instead of finding the intrinsic reward for doing the right thing on their own. This is probably extensively researched already, although i don't have sources handy right now.
2) Along time, positive reinforcements behave in accordance with the "law of diminishing returns". i.e. If you always reward positive actions with some standard reward, it will lose effect over time. With the danger of the positive action not feeling worthwhile anymore.
I thus became a great believer in the importance of negative consequences for bad behavior. I mostly apply (within reason) Shame and Restriction of freedoms to punish my kids for bad behavior (reserving the occasional physical slap for very serious offenses). Shame is a motivator that interests me a lot as a parent. I think it is an essential motivator in society in general, but the most delicate one. Too much Shame can also make you socially inept.
In the end I think its always about the right balance...
You can join them here: https://discord.gg/S9uJ794
https://m.youtube.com/watch?v=NcZCzr9ExKk
Also see this one for extra footage of the Gyroless, "trail-less" TMS bicycle that still manages to balance itslef
But they don't scale up with numbers or across geographically. The scary thing about automated processes of facial recognition is precisely the ability to scale to everyone and everywhere.
My goal was get the paper plane to catch the updrafts and actually climb up from my 3rd floor to the top of the building. It was very rare, I must have tried 10s of different fold types but sometimes conditions would be just perfect and a particular plane would ride up in circles sometimes even clearing the top floor of the building and then glide down and land sometimes hundreds of meters away or even vanish from view.
When I ran out of planes I would grab my bicycle and go out to recover as many of them possible, come back up and repeat...
...sweet times =)
...i started wondering about this when I was reading a few articles about the AlphaZero algorithm that learned to play chess entirely from self play and wondered if it would always play the same moves in response to a fixed set of opponent moves (assuming the opponent starts as white). My guess was that it wouldn't always respond in exactly the same way in case there's any MCTS like step somewhere in there blended with the Machine learning algorithm.
For a game like chess it would seem to make sense that the overall algorithm would still include a MCTS step (like AlphaGo did) but for an autonomous car it would seem crazy to any human to imagine that there would be any random search for a decision in a tree of possible interpretations of the input for example.
Does any one have any detailed knowledge about this? Would a non-deterministic algorithm ever be allowed in an autonomous car?
So, if someone out there can explain, is a program like AlphaGo deterministic? Meaning if the opponent plays the exact same moves in 2 consecutive games will the 2nd game be a replica of the first?
My naive guess is that the MTC search part of the algorithm will necessarily introduce some variation in play, no?
Is this really true? Would a smartphone chess app beat the worlds top ranked player today?
https://insideevs.com/musk-says-biggest-model-3-bottleneck/
> “To date, our primary production constraint has been in the battery module assembly line at Gigafactory 1, where cells are packaged into modules.” “We had to rewrite all of the software, from scratch. We managed to write 20 to 30 man-years of software in 4 weeks.”
I'm an amateur programmer (Python, JS) and I have been wondering for a long time if there is a language more readable than python.