AI may be the next revolutionary technology and have the impact comparable to Neolithic and Industrial revolutions

1️⃣ to discuss the impact of AI it is useful to situate it within two “reference classes” of technologies: general purpose technologies (GPTs) and revolutionary technologies

• GPTs examples: electricity and the digital computer, revolutionary technologies examples: domesticated crops (Neolithic Revolution) and steam engine (Industrial Revolution)
• GPTs are distinguished by their unusually pervasive use and large inherent potential for technical improvement, it normally takes several decades for them to achieve large-scale impacts
• a revolutionary technology is a GPT that supports an especially fundamental transformation in the nature of economic production

2️⃣ AI may eventually emerge as another revolutionary technology, by drastically reducing the role of human labor in economic production

• revolutionary technologies produce changes that are far more fundamental and far-reaching than anything policymakers have experienced
• Neolithic Revolution: larger, more complex, and more hierarchical social institutions; declined average living standards, levels of gender equality, and levels of violence
• Industrial Revolution: growth in the average person’s wealth; new trend toward more widespread democracy; rising lifespans, education, between-region inequality, gender egalitarianism; declining rates of slavery

3️⃣ as for now, the impact of AI has been limited by technical bottlenecks & implementation challenges

• technical bottlenecks: existing techniques, sources of data, and quantities of computing power are not sufficient
• implementation challenges: scarcity of expertise, regulatory barriers, the unavoidable complexity of many engineering projects, the need to invent complementary technologies, and the need to attract large investments of capital

4️⃣ if researchers can make enough progress on relevant technical bottlenecks, then the technology’s potential for long-term improvement is enormous

• AI may be able to perform all of the tasks that people are capable of: this milestone will probably be reached within the current century
• potential employers have little reason to hire humans if it would always be cheaper and at least as effective to use a machine
• existing AI techniques could raise income inequality or unemployment but redistributive policies could reverse this apparent consequence
• in the political domain: more effective law enforcement, political persuasion, and video forgery; reduced crime or risk from terrorism, due to law enforcement applications, or improved government decision-making due to better data-driven analysis
• AI systems could largely replace human governments in their roles providing services, enforcing rules, and making various decisions about how best to use resources

📰The Impact of Artificial Intelligence: A Historical Perspective - Ben Garfinkel, Governance of AI Program Researcher https://docs.google.com/document/d/1I13_0o3kUe1AVQNfevOF9sHpc4mCQkuFDxOXFj_4g-I/edit#