Ten essential observations
The Singularity Principles arise from a number of general observations.
1. Greater tech power enables more powerful results
The more powerful technology becomes, the more powerful are the results it can produce – including powerfully good results and powerfully bad results. Small firecrackers can, in some cases, ignite larger fires and cause widespread damage. But larger explosives are more likely to destroy an entire city in the blink of an eye.
2. Different perspectives evaluate “good” vs. “bad” differently
Results that are evaluated as “good” from one perspective, such as an increase in profits or in market share, or breaking records for speed or performance, can also be evaluated as “bad” from other perspectives – for example when externalities are factored into calculations, or when a broader view of human flourishing is considered.
3. Tech breakthroughs are unpredictable in both timing and impact
Breakthroughs in technological capability may be dramatic and unexpected. That’s both in terms of timing (such as a sudden breakthrough following a long period of slow progress) and in terms of impact (with some breakthroughs being more widely applicable than was previously anticipated).
4. Potential complex interactions make prediction even harder
Complex interactions between multiple developments in different fields – including disruptive breakthroughs as well as incremental progress, and including changes in legal systems or philosophical outlooks, as well as diverse overlapping technological changes – make outcomes even harder to anticipate and predict. That’s why we have to work harder, and smarter, at the tasks of anticipation and surveillance.
5. Some tech failures would be so drastic as to rule out recovery
Although there are many technological failures from which it’s possible to recover, with people growing wiser (hopefully) as a result, some technological failures may have such a large scale that subsequent recovery would be extremely hard or even impossible. In these cases, there’s no hope for “failing forward” or “failing smart”. The only option in these cases is to avoid the failure in the first place. The more powerful the underlying technology, the more attention needs to be paid to such possibilities.
6. A past string of good results is no guarantee of future wellbeing
The mere fact that a piece of technology has delivered a string of good results in the past does not guarantee, by itself, that the technology in question will deliver good results in altered circumstances in the future. The management of technological change therefore needs to rely on more than induction, that is, more than the assumption that the future will continue to resemble the past.
7. It’s insufficient to rely on good intentions
Something else that it’s insufficient to rely on is the perceived good intentions of individuals or companies – intentions that these individuals or companies will avoid any very bad outcomes. Alas, when good intentions are coupled with a mistaken or incomplete understanding of an issue, they can result in the very sort of bad outcomes they were seeking to avoid. Moreover, these good intentions can sometimes become submerged under other forces that are more powerful.
8. Wishful thinking predisposes a blindness to problematic issues
Due to wishful thinking, providers of potential new technological solutions are often inclined to turn a blind eye to problematic features that may arise. If they hear reports of adverse side-effects or possible unintended consequences, they are motivated to disbelieve these reports, or to distort them or throw doubt on them.
9. Competition can be hazardous as well as beneficial
Although a competitive marketplace can often accelerate progress, with companies racing to discover and apply new innovations, such competition can also result in dangerous corner-cutting or other reckless risk-taking. Hostile arms races are particularly hazardous.
10. Results can be transformed if human attributes are transformed
In anticipating potential scenarios, there’s no need to regard human institutions, human attitudes, and human intentions as fixed and unchangeable. Changes in all these aspects of the human outlook can be part of wise responses to technological risks and opportunities. On the other hand, changes in these human attributes can also make matters worse. That’s why a truly broad perspective is needed.
In summary, the above observations make it essential that we improve our abilities to anticipate and manage disruptive change.
These observations provide the basis for the Singularity Principles.
Note: The following video from the Vital Syllabus contains a visual illustration of these ten essential observations. (In some cases, the description of these observations has evolved since the video was originally recorded.)
Postscript: What about ethics?
The development and deployment of technologies is sometimes addressed from the viewpoint of ethics. Thus it is common to hear advocacy for “the ethical use of technology”, or, more simply, for “AI ethics”.
This framing sometimes causes problems. Ethicists are sometimes perceived as people who mainly say “no”, as in “thou shalt not xxx” and “thou shalt not yyy”.
Moreover, the common ethical injunctions to ensure “fair” or “equitable” access to technologies are subject to controversy, since there are divergent views on what counts as “fair” and “equitable”.
The Singularity Principles take as their starting point, not any appeal to ethics, but a more basic set of considerations:
- Decreasing the probability of severe harm
- Increasing the probability of profound benefit
Nevertheless, the principles endorse one of the fundamental insights of ethics:
- Just because we believe we could develop some technology, and even if we have some desires to develop that technology, that’s not a sufficient reason for us actually to go ahead and develop it and deploy it.
More briefly: could does not imply should.
The underlying reasons to avoid developing or deploying some technology, despite appeals in favour, are that the technology
- Would be likely to increase the probability of severe harm, or
- Would be likely to decrease the probability of profound benefit.