Differences

This shows you the differences between two versions of the page.

--- arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:start [2023/03/09 04:54]
katjagrace
+++ arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:start [2024/05/28 19:23] (current)
katjagrace ↷ Links adapted because of a move operation
@@ Line 1: / Line 1: @@
 ====== Is AI an existential risk to humanity? ======
-//This page is under active work and may be updated soon.//
+//This page is under active work and may currently be incoherent or inaccurate.//
-The balance of evidence appears to suggest that AI poses a substantial existential risk, though none of the arguments that we know of appear be strongly compelling evidence.
+The balance of evidence appears to suggest that AI poses a substantial existential risk, though none of the arguments that we know of appear be conclusive evidence.
 ===== Background =====
 Many [[arguments_for_ai_risk:list_of_sources_arguing_for_existential_risk_from_ai|thinkers]] believe advanced [[clarifying_concepts:artificial_intelligence|artificial intelligence]] (AI) poses a large threat to humanity's long term survival or flourishing. Here we review evidence.
+For views of specific people working on AI, see [[arguments_for_ai_risk:views_of_ai_developers_on_risk_from_ai|this page]].
 Note that arguments included here are not intended to be straightforwardly independent lines of evidence. They may instead represent different ways of conceptualizing and reasoning about the same underlying situation.
-===== Scenarios and supporting arguments =====
+===== Arguments =====
+//(Main article: [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:scenario_malign_ai_agents_control_the_future:start|Will malign AI agents control the future?]])
+//
+[[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:list_of_arguments_that_AI_poses_an_xrisk|Several arguments]] have been made for expecting artificial intelligence to pose an existential risk. The [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:will_malign_ai_agents_control_the_future:argument_for_ai_x-risk_from_competent_malign_agents:start|most prominent argument]] for AI posing a severe threat to humanity is for a [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:scenario_malign_ai_agents_control_the_future:start|scenario where competent, malign agents control the future]], and can be summarized briefly as follows:
+  - Some advanced AI systems will very likely be 'goal-oriented'
+  - The aggregate goals of these systems may be bad. (There are reasons to think this.)
+  - Such systems will likely have the power to achieve their goals even against the will of humans
+  - Thus, there is some chance that the future will proceed in opposition to long-run human welfare, because these advanced AI systems will succeed in their (bad) goals
+===== Other arguments =====
-Advanced AI could conceivably threaten humanity's future via several different disaster scenarios, each of which is suggested by different arguments. We consider these scenarios and arguments here, one at a time.
+Further arguments for AI posing an existential risk to humanity can be categorized by scenario.
 ==== Scenario: malign AI agents control the future ====
-//(Main article: [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity: will_malign_ai_agents_control_the_future:start|Will malign AI agents control the future?]])
+//(Main article: [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:scenario_malign_ai_agents_control_the_future:start|Will malign AI agents control the future?]])
 //
-This appears to be the most discussed extinction scenario. In it:
+Arguments that this scenario will occur include:
-  - AI systems are created which, a) have goals, and b) are each more capable than a human at many economically valuable tasks, including strategic decision making.
+  * **AI will replace humans as most intelligent 'species'** \\  \\ //(Main article: [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:will_malign_ai_agents_control_the_future:argument_for_ai_x-risk_from_most_intelligent_species|Argument for AI x-risk from most intelligent species]])// \\  \\ **Summary**: Humans' dominance over other species in controlling the world is due primarily to our superior cognitive abilities. If another 'species' with better cognitive abilities appeared, we should then expect humans to lose control over the future and therefore for the future to lose its value.  \\  \\ **Apparent status**: Somewhat suggestive, though doesn't appear to be valid, since intelligence in animals doesn't appear to generally relate to dominance. A valid version may be possible to construct.
-  - These AI systems' superior performance allows them to take control of the future, for instance through accruing social and economic power, or through immediately devising a plan for destroying humanity.
-  - The AI systems do not want the same things as humans, so will bring about a future that humans would disprefer
-This scenario includes sub-scenarios where the above process happens fast or slow, or involves different kinds of agents, or different specific routes, etc.
+  * **AI agents will cause humans to 'lose control'** \\  \\ **Summary**: AI will ultimately be much faster and more competent than humans, so either, a) must make most decisions because waiting for humans will be so costly, b) will make decisions if it wants, since humans will be so relatively powerless, due to their intellectual inferiority. Losing control of the future isn't necessarily bad, but is prima facie a very bad sign. \\  \\ **Apparent status**: Suggestive, but as stated does not appear to be valid. For instance, humans do not generally seem to become disempowered by possession of software that is far superior to them.
-Various [[https://wiki.aiimpacts.org/doku.php?id=arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:will_malign_ai_agents_control_the_future:start#arguments|arguments]] are made for this scenario. The most prominent appears to be:
+  * **Argument for loss of control from extreme speed** \\ \\ **Summary**: Advancing AI will tend to produce very rapid changes, either because of feedback loops in automation of automation processes, or because automation tends to be faster than the human activity it replaces. Faster change reduces human ability to steer a situation, e.g. reviewing and understanding it, responding to problems as they appear, preparing. In the extreme, the pace of socially relevant events could become so fast as to exclude human participation. \\  \\ **Apparent status**: Heuristically suggestive, however the burden of proof should arguably be high for an implausible event such as the destruction of humanity. This argument also seems to support concern about a wide range of technologies, which may be correct.
-  * **AI developments will produce powerful agents with undesirable goals** \\  \\ //(Main article: [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:will_malign_ai_agents_control_the_future:argument_for_ai_x-risk_from_competent_malign_agents:start|Argument for AI X-risk from competent malign agents]])// \\  \\ **Summary**: At least some advanced AI systems will probably be 'goal-oriented', a powerful force in the world, and their goals will probably be bad by human lights. Powerful goal-oriented agents tend to achieve their goals. \\  \\ **Apparent status**: This seems to us the most suggestive argument, though not watertight. This seems prima facie plausible, but destroying everything is a very implausible event, so the burden of proof is high.
+In light of these arguments, this scenario seems to us plausible but not guaranteed. Its likelihood appears to depend strongly on the strength of one's prior probability on arbitrary risks being sufficient to destroy the world.
-In light of [[https://wiki.aiimpacts.org/doku.php?id=arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:will_malign_ai_agents_control_the_future:start#arguments|arguments]], this scenario seems plausible but not guaranteed.
 ==== Scenario: AI empowers bad human actors ====
@@ Line 56: / Line 66: @@
 Competition can produce outcomes undesirable to all parties, through selection pressure for the success of any behavior that survives well. AI may increase the intensity of relevant competitions.
-===== General evidence =====
+==== General evidence ====
 This is evidence for existential risk from AI which doesn't point to specific scenarios:
@@ Line 82: / Line 92: @@
   - AI performance may increase very fast due to inherent propensities to discontinuity
   - AI performance may increase very fast once AI contributes to AI progress, due to a feedback dynamic ('intelligence explosion' from 'recursive self improvement')
+There are various other [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:scenarios_and_arguments|scenarios and supporting arguments]]
+====== Conclusion ======
+In light of [[arguments_for_ai_risk:is_ai_an_existential_threat_to_humanity:scenario_malign_ai_agents_control_the_future:start|these arguments]], this scenario seems plausible but not guaranteed.
 ====== See also ======
@@ Line 89: / Line 108: @@
   * [[arguments_for_ai_risk:list_of_sources_arguing_for_existential_risk_from_ai|List of sources arguing for existential risk from AI]]
   * [[will_superhuman_ai_be_created:start|Will superhuman AI be created?]]
+  * [[arguments_for_ai_risk:list_of_possible_risks_from_ai|List of possible risks from AI]]
 ====== Notes ======

AI Impacts Wiki

User Tools

Site Tools

Differences

Page Tools