Interestingly, many body parts have 2-3 different functions despite modularity. A mouth could be used for drinking, eating, biting, speaking and breathings; legs – for running and fighting
avturchin
Karma: 1,284
Will blackboxing the reward function help, either physically or cryptographically? It also should include the obscurity about the boundary between the BB and internal computations in AI, that is, the AI will not know which data actually trigger the BB reaction.
This is how human reward function seems to work. It is well protected from internal hacking: if I imagine that I got 100 USD, it will not create as much pleasure as in the situation when I am actually getting 100. When I send mental image of 100 USD into the my reward box, the box “knows” that I am lying and don’t generate the reward. As don’t know much about how the real human reward function works I have to get real 100 USD.
In October 2019, a model was trained by Google with on 750 GB training data and it has 11 billion parameters (vs. 40 Gb and 1.6B for GPT-2 8 months before that.)
I could use a fair coin to decide should I open the envelope. In that case I become unpredictable.
If the universe is infinite and has all possible things, then most of ignorance becomes randomness?
Assuming that we have no less than 20 problems and for each problem we have 80 per cent chances of success (if we know more, it is not a problem) we have total only 1 per cent of the total probability of success.
So, this method produces very pessimistic expectations even if problems themselves seems solvable. EY wrote somewhere that multiplying probabilities is bad way to estimate the chances of success of cryonics, as this method underestimate the growth of experience of the problem solver.
Another takeaway could be that we should search total AI safety solutions where we have less unknowns.
One wrong take on “taking over the world” is “having causal power to change everything”. The reason for it is that because of the “butterfly effect” every my action will change fates of all future people, however, in a completely unknown way.
“Normative assumptions” by Stuart Armstrong discussion seems relevant here.
My personal estimate is 10 per cent in 10 years. If it is distributed linearly, it is around 0.2 per cent until the end of 2019, most likely from unknown secret project.
I am bad visual thinker, but I was able to reach much higher performance on dual n back after I found the trick to write down all numbers on an imaginary board.
For me, it is evidence for AGI, as it says that we only just one step, may be even one idea, behind it: we need to solve “genuine causal reasoning”. Something like “train a neural net to recognise patterns in in AI’s plans, corresponding to some strategic principles”.
Dating is over-advertise as an effective way of getting a GF, as it helps cafes, flowers sellers, girls themselves etc. In my experience, I’ve got GFs in two ways:
1) Relation evolved from friendship on a course of years.
2) I lived my own life, a girl fall in love at me based her own internal processes, she showed interest, I responded and we started dating.
So, the secret is: live your own interesting life, and if a girl falls in love at you, start dating her. But stopped chasing these 9th and 10th!
My experience: dating is just a wrong way to come to relationship. I have had several relationships and never through proper dating. I read “Mate” and it didn’t help.
Yes, they had Hungarian-Jewish ancestry which is known to produce many genius minds called the Martians.
It is interesting if quantum computers’ computations could be used as evidence against “we are in simulation on a classical computer”. Quantum supremacy can’t be modeled on a classical computer. But the simulators could easily overcome the obstacle by using quantum computers too. However, it creates some limits about the nature of their hardware.
One could imagine an experiment to test if we are in a classical simulation by performing a very complex computation of a quantum computer. But this could also end our world.
This was tested in the chess field. A family decided to teach their kids chess from early childhood and created three grandmasters (Plogar sisters) https://www.psychologytoday.com/intl/articles/200507/the-grandmaster-experiment (But not one Kasparov, so some variability of personal capabilities obviously exists.)
Another way to look at human intelligence is to assume that it is a property of the whole humanity, but much less of individual humans. If Einstein was born in a forrest as a feral child, he will not be able to demonstrate much of his intelligence. Only because he got special training he was able to built his special relativity theory as incrementation of already existing theories.
There is no proved working antiaging technologies. Some possible interventions are relatively cheap, e.g. metfomin bulk cost is few dollars for kg, but they could provide at best 1-3 years of life expectancy.
Given the exponential decline of the price of any new tech, future anti-aging tech (e.g. nanobots-based vaccine against aging) will be as cheap as a new smartphone eventually. For example, other current life-saving drugs are also very cheap now: vitamins, vaccines, antibiotics.
Thus saving life will be cheaper than rising a new person (at least until mind uploading when copying will be cheaper).
Other possible success stories are semi-success stories, where the outcome is not very good, but some humans survive and significant part of human values is preserved.
One case of the semi-success story is that many sovereign AIs control different countries or territories and implement different values in them. In some of these territories AIs’ values will be very close to the best possible implementation of aligned AI. Other AIs could be completely inhuman. Slow takeoff could end in such world of many AIs.
Another case is that unfriendly AI decides not to kill humans for some instrumental reasons (research, acausal trade with other AI, just not bother killing them). It could even run many simulations of human history including simulations of friendly AIs-sovereigns and their human civilizations. In that case, many people will live very happy lives despite being controlled by unfriendly AI. Like some people were happy under Saddam Hussein rule.
Semi-success stories could be seen as a more natural outcome as we typically don’t have perfect things in life.
A friend told me this story many years ago: He was working on repairing some electronic staff and one block had a green light turned on when the block was turned off by a switch, but not disconnected from the power line. However, there are no short circuits in it. After long investigation, mostly for curiosity, he found that some piece of alloy covered another piece and together they created a capacitor, which was able to let in AC part of incoming signal and power the light in the gadget.