Fast Takes: The real route to AI safety, conflict is inevitable and more

AI safety may be “less a problem of ethics” than of “engineering,” suggests The Washington Post Editorial Board.“Deterministic software that exists outside the AI to enforce the rules” can “place hard limits” on what AI agents can do.

Just as engineers in the “early web” days “hardened the environment” software operated in by “granting it only minimal access rights,” AI agents “should be treated the same way.” No matter “how well aligned an agent becomes” with human intentions, “its ability to act should always be limited by systems fully outside its control.” Ultimately, the “machines don’t have to be trustworthy if the systems around them are.”The “quest for a stable global order has long beguiled idealists and realists alike,” but “conflict is inevitable,” argue Tod Lindberg & Corban Teague at The Wall Street Journal, so global “stability” is “at best an aspiration” and more likely “happy talk.” Times like the “post-Cold War period of general peace” are at “variance with the normal state of international affairs — violence, conflict and disorder.” Now we’re entering “a new era” of uncertainty, where America faces “not merely competitors, challengers or even adversaries, but enemies” — a truth that our enemies understand far better than we do.“Debates or spheres we view as mere competition” to the other side are “economic or informational warfare.”The “rising tide” of unaffordability “is lifting Democrats’ prospects this fall,” David DesRosiers notes at RealClearEnergy, yet “Democratic policy is not prescriptive to mitigating that pain.

It only makes it worse.” Fact is, “The rising tide of general prosperity is locked at the hip to an abundant supply of affordable energy,” and “Republicans fixate on juicing supply, Democrats on regulating demand.The former brings more supply and lower costs; the latter constrains supply and drives prices up.�...