Perhaps a followup study can investigate if trying to sneak culture war topics into ostensibly non-political spaces also maps to dark triad.
Capybasilisk
Karma: 74
I guess you could posit natural selection as being objective reality’s value system, but I have the feeling that’s not the kind of thing moral realists have in mind.
I also recommend Roman Yampolskiy’s article “The Space of Possible Mind Designs”.
From the point of view of the alien scientists, I’d argue that it’s rational to ignore claims of unprecedented insight into reality unless the claimant can demonstrate impressive and unexplained feats using said insight.
I would think that gaining an entirely new sensory modality would lead to unparalleled advantages over fellow members of your species. At the very least, it would let you do things that would confound them.
New colors.
I don’t mean discerning ever finer gradations in existing colors, but entirely new color qualia.
What would it be like? Would our brains be able to integrate this new phenomenal experience?
And most importantly, if color is a property of brains rather than something in the external world, does that imply the number of “possible colors” is infinite? I.e. seeing as biological brains “choose” how to internally represent a particular wavelength of EM radiation, is the seemingly Platonic realm from which these colors are plucked from inexhaustible?
We currently have aesthetic preferences over existing colors. I would want to know whether these possible colors can be enumerated, searched through, and have a utility function placed over them so that each human can find what is to them “the best possible color.”
The Arbital entry on Unforeseen Maximums [0] says:
“Juergen Schmidhuber of IDSIA, during the 2009 Singularity Summit, gave a talk proposing that the best and most moral utility function for an AI was the gain in compression of sensory data over time. Schmidhuber gave examples of valuable behaviors he thought this would motivate, like doing science and understanding the universe, or the construction of art and highly aesthetic objects.
Yudkowsky in Q&A suggested that this utility function would instead motivate the construction of external objects that would internally generate random cryptographic secrets, encrypt highly regular streams of 1s and 0s, and then reveal the cryptographic secrets to the AI.”
[0] https://arbital.greaterwrong.com/p/unforeseen_maximum/
Thanks for sharing this.
Would there be any advantages to substituting brute force search with metaheuristic algorithms like Ant Colony Optimization?
Of possible interest, Roman Yampolskiy’s paper, “The Universe Of Minds”.
https://arxiv.org/pdf/1410.0369.pdf
The paper attempts to describe the space of possible mind designs by first equating all minds to software. Next it proves some interesting properties of the mind design space such as infinitude of minds, size and representation complexity of minds. A survey of mind design taxonomies is followed by a proposal for a new field of investigation devoted to study of minds, intellectology. A list of open problems for this new field is presented.
Are random trading strategies more successful than technical ones?
In this paper we explore the specific role of randomness in financial markets, inspired by the beneficial role of noise in many physical systems and in previous applications to complex socioeconomic systems. After a short introduction, we study the performance of some of the most used trading strategies in predicting the dynamics of financial markets for different international stock exchange indexes, with the goal of comparing them with the performance of a completely random strategy. In this respect, historical data for FTSE-UK, FTSE-MIB, DAX, and S&P500 indexes are taken into account for a period of about 15-20 years (since their creation until today).
...
Our main result, which is independent of the market considered, is that standard trading strategies and their algorithms, based on the past history of the time series, although have occasionally the chance to be successful inside small temporal windows,on a large temporal scale perform on average not better than the purely random strategy, which, on the other hand, is also much less volatile.In this respect, for the individual trader, a purely random strategy represents a costless alternative to expensive professional financial consulting, being at the same time also much less risky, if compared to the other trading strategies.
But a lot of that feeling depends on which animal’s insides you’re looking at.
A closely related mammal’s internal structure is a lot more intuitive to us than, say, an oyster or a jellyfish.
For example, take the idea that an AI should maximise “complexity”. This comes, I believe, from the fact that, in our current world, the category of “is complex” and “is valuable to humans” match up a lot.
The Arbital entry on Unforeseen Maximums elaborates on this:
Juergen Schmidhuber of IDSIA, during the 2009 Singularity Summit, gave a talk proposing that the best and most moral utility function for an AI was the gain in compression of sensory data over time. Schmidhuber gave examples of valuable behaviors he thought this would motivate, like doing science and understanding the universe, or the construction of art and highly aesthetic objects.
Yudkowsky in Q&A suggested that this utility function would instead motivate the construction of external objects that would internally generate random cryptographic secrets, encrypt highly regular streams of 1s and 0s, and then reveal the cryptographic secrets to the AI.
Robin Hanson posits that the reason why there isn’t wider adoption of prediction markets is because they are a threat to the authority of existing executives.
Before we reach for conspiracies, maybe we should investigate just how effective prediction markets actually are. I’m generally skeptical of arguments in the mold of “My pet project x isn’t being implemented due to the influence of shadowy interest group y.”
As someone unfamiliar with the field, are there any good studies on the effectiveness of PM?
This would just greatly increase the amount of credentialism in academia.
I.e., unless you’re affiliated with some highly elite institution or renowned scholar, no one’s even gonna look at your paper.
Look on the bright side. If it turns out to be a disaster of Black Death proportions, the survivors will be in a much stronger bargaining position in a post-plague labour market.
Consider the trilobites. If there had been a trilobite-Friendly AI using CEV, invincible articulated shells would comb carpets of wet muck with the highest nutrient density possible within the laws of physics, across worlds orbiting every star in the sky. If there had been a trilobite-engineered AI going by 100% satisfaction of all historical trilobites, then trilobites would live long, healthy lives in a safe environment of adequate size, and the cambrian explosion (or something like it) would have proceeded without them.
Altman’s responsiveness test might be more useful if it measured a founder’s reply time to an underling rather than to the guy holding the purse strings/ contact book.
An alternative hypothesis is that Graham has a bunch of tests/metrics that he uses to evaluate people on, and those tests/metrics work much better when people do not try to optimize for doing well on them
Isn’t it a bit odd that PG’s secret filters have the exact same output as those of staid, old, non-disruptive industrialists?
I.e., strongly optimizing for passing the tests of <Whitebread Funding Group> doesn’t seem to hurt you on YC’s metrics.
Relatedly, is there any work being done to come up with distance measures for agent states (including any of momentary experiences, life-sums, or values)? We talk a lot about marginal and absolute value of lives and make comparisons of experiences (motes vs torture) - what, if anything, has been done to start quantifying?
It turns out that measuring the distance between possible worlds is...complicated.
What about when the poor men come barging through the rich man’s gate? I take it that too is factored into the divine plan?
Leo Gao thinks it might be possible to ride language models all the way to AGI (or something reasonably close):
https://bmk.sh/2021/06/02/Thoughts-on-the-Alignment-Implications-of-Scaling-Language-Models/