Sunday, April 29, 2007

Capitalist or Socialist Upload Scenario?

There are at least two scenarios for how mass uploading could occur, the Capitalist model and the Socialist model.

1. Capitalist Model (Evolutionary)
In the Capitalist model, uploaders would choose from a selection of storage, processing and security packages. Offerings would be presented with marketing aplomb, perhaps featuring modules analogous to the Gandhi [minimalist], the Toyota [adequate, fully functional] and the Cadillac [premium]. The capitalist model is a Darwinist or evolutionary approach.

2. Socialist Model (Developmental)
In the Socialist model, all storage, processing and security upload packages would be the same, perhaps directed by some sort of quasi or overtly governmental or private advocacy body. Each uploader would receive the same storage, processing and security resource module.

At first blush, the Capitalist model seems most likely; it is a logical extension of how economics and marketplaces function in the current world. However, the key issue is the hyper-evolution capability that is assumed to occur in the digital substrate. Perhaps uploads should at least be equalized at the beginning such that all those who upload have the same chance of becoming a super-intelligence.

However, trying to equalize uploads (such as via a handicapping system) or the resources available to them at the offset is both unlikely to occur, and more importantly, unlikely to make a difference. The magnitude of digital evolutionary changes makes the starting point irrelevant, both in terms of capability and processing power.

The real question is what processing power will become available to digital intelligences post-upload or post-creation and the resulting evolution and goals which may become at odds with those of biological humanity. It is not reasonable to assume that external control can be imposed for long.

Sunday, April 22, 2007

Historical Simulation Ethical Issues

Historical Sims will be a tremendously useful tool but they are also rife with ethical issues, even without considering that ethics will also evolve morefold by the time historical sims are a norm.

Sim Influencing and Harvesting
To what degree (if any) will external influencing, probing, shifting, etc. be acceptable once the sim has begun? Outsiders do not bring medicine and electricity to tribes living on Earth, the general thought is not to interfere with less-technologically advanced cultures if they are found elsewhere in the world or the universe, but many might want future technology introduced from outside if it were possible. One of the great potential benefits of a sim would be to extract and deploy any developed technologies that do not exist; is this ethical? Both the technology extraction and the introduction would need to be considered, for example, should attribution and royalties accrue to the sim world?

Sim Participant Ethics
Perhaps the largest class of ethical issues relates to sim participants. Self-aware agents (either recruited or created) as sim participants will presumably be more effective than non self-aware agents. The sim participants will either know that they are in a sim and have accepted the historical terms (as in Charles Stross' Glasshouse), particularly the primitive pain and misery of historical harm and death (war, disease, etc.), or not know that they are in a sim, in which case perhaps some argument could be made that it is all right that they do not know since in fact we cannot prove that we are not in a sim ourselves.

Sim Mindfile Access and Participant Privacy
One benefit of sim space is that not just the event outcomes can be known but also much more about the views and experiences of all of the participants not just those who wrote about it or were written about and even before writing and records were preserved. Is it appropriate and ethical to "mindtap" without the sim participant's permission? If the participants know they are being mindtapped, will that heisenberg behavior? Although lifecasting is rapidly approaching normalcy in the current technophilic world, it would be culturally unacceptable in historical periods.

Sim Termination
The issues concerning the termination of sims are non-trivial. Given unlimited processing power, and as a test of existential risk, there is an argument to let all sims continue indefinitely. How is extinguishing a sim not the murder of the self-aware participants or will it be desirable and possible to offer mindfile backups to all participants? This can be tricky, should "death rights" be different for those dying naturally within the sim vs. those dying in an extinguishment of the sim? What are the best approaches if the sim culture does not yet have mindfile backup technology; introduce the idea to sim participants and let them choose, save everyone irrespectively (this could be the future law anyway) with or without their knowledge or allow extinguishment? What would the options be for a mindfile exiting a sim? A state-assigned runtime area and survival resources, and access to the dial-a-world console?

Sunday, April 15, 2007

Sims Sponsor Technological Advancement

Running simulations of the past, present and future will be valuable tools in exploring, understanding and testing any and every situation.

What if Rome did not fall, what if Germany won WWII, what if humans were not corrupt, what if it were socially unacceptable for educated women not to work in the paid labor force, what if hormone management programs had become possible and de rigueur centuries ago, what if the Pangaea split had been different, what if the Chicxulub asteroid had not rendered the dinosaurs extinct, what if sustainable life had started earlier?

Technological Advancement
The most useful sims would be those either arising naturally or "managed" such that advancement optima were achieved and the sim world evolved at a faster rate than the world that initiated the sim (how many sims and iterations will that take?) Sims could be run at sped-up time to iterate and generate results more quickly. Ideas and technologies could then be harvested for use in the initiating world. (Would that be ethical? Would that be cheating?)

It might be easier and more desirable to join the sim world, at least with a self-copy rather than extract and apply the technology. Lequel est plus réelle? Some interesting questions could be studied regarding technology relevance in the absence of the ideology, politics, economics, culture [and world] of its founding.

The current human population is gated by physical-world resource constraints, sims could make it possible for billions, trillions, etc. of 'people,' self-aware autonomous intelligent agents, to exist. Would this be desirable? Would this extend human knowledge, productivity, creativity and happiness? What kind of restrictions, fears, threats and opportunities would this impose?

Sunday, April 08, 2007

Prosper's $50m loan pool - high risk, high reward?

Prosper has reached a milestone of $50 million in P2P loans extended, however $10 million of these may end in default. Despite the high rate of Prosper defaults, much higher than traditional consumer credit defaults, lenders should not and may not mind if they are receiving appropriate returns.

P2P lending is emerging as a new credit category, not just in the visible ways of loan origination and delivery but also in the financial sense of how risk and reward are defined. P2P lenders are able to accept higher default rates since they are also theoretically realizing higher returns. Some portion of the 10% traditional spread in bank lending between borrowing and lending accrues to the lender.

It is clear that Prosper loans default at higher rates than traditional unsecured consumer credit loans. The chart below shows Prosper defaults in pink and Experian (as a proxy for the consumer credit market as a whole) defaults in blue. In every credit tier, Prosper loans have higher defaults. In the prime market of AA, A, B and C credit tiers, Prosper narrowly underperforms Experian. However as credit quality worsens, so do Prosper defaults with Prosper loans defaulting at double traditional rates in the E and HR (high risk) tiers.

Default data can be found at the Prosper website by scrolling to the bottom of the Performance page and selecting the Estimated ROI link. Lender ROI estimates are trickier, ranging from a blended portfolio ROI of -1% using the Prosper site data to 17% using the data from Eric's Credit Community.

The marketplace aspect of Prosper is working as sub-prime borrowers have seen the opportunity and are creating most of the volume on the site, 75% of listings and 60% of fundings. Although many loans receive funding that probably would not in traditional credit settings, the majority (75%+) of listings do not get funded.

Prosper is only about a year old and the P2P lending market needs to achieve much higher volumes before meaningful performance can be evaluated. The question is whether rates of return can be delivered which are appropriate given the higher risk from the higher defaults and if lenders can learn how to price default risk effectively in this new credit product.

Thursday, April 05, 2007

My house is an AI

Smart homes sound great but what about the next step, what will it be like when dwellings are artificial intelligences (AIs)?

At first, tools that know people better than they know themselves including predicting behavior and speech may be disconcerting and even creepy and people might not want to live in a dwelling that is an AI. However after getting used to it, presumably there will be many benefits, including continuous health monitoring, companionship, critical feedback and transfer of menial tasks. The AI would be residence based with mobile accompaniment.

Residents will likely be able to select a personality interface for their AI, or even multiple personality interfaces. The critical feedback personal AI would be a nice complement to the socially groomed empathy received from friends and family. Deception and excuses would be nearly impossible "I felt sick all day" might elicit a response like "No, your body status and health levels were fine but you watched ShowStash content for 7 hours." A friendly diplomatic upbeat personality interface would be a good default for the 'inner critic' team member, rather than a HAL-like interface.

A genial AI interface, "the buddy," would be another logical member of one's AI personality suite. This personality could provide companionship and monitoring for everyone, particularly higher needs-based individuals like seniors and children. Some might worry that a personal AI could become a more fulfilling emotional companion than a human partner but most technologies mean more not less, and more ways an individual could be emotionally fulfilled would be a good thing.

New industries could spring up, for example AI personality interface design and human-AI interaction psychology, including a new kind of dysfunctional home relationship.

Residence AIs would be autonomous entities licensed by local municipalities. A person would negotiate rent and other aspects directly with the AI. The AI would have purchasing authority and run the household, essentially acting as a more resource appropriate implementation of the concept of wife.