Most Replayed Moment: AI Safety Expert Predicts The Next 20 Years! Will It Really Take All Jobs?

May 22, 2026
Overview

Dr. Roman Yampolskiy, an AI safety expert, details how superintelligence will automate all jobs, making traditional retraining obsolete. He emphasizes AI safety as the paramount existential risk, highlighting the unpredictability of advanced AI and the futility of assuming human control.

At a Glance
9 Insights
36m 29s Duration

Deep Dive Analysis

1. Don’t Dwell on Unpreventable Doom

Adopt a mindset of not dwelling on catastrophic outcomes that cannot be modified, similar to how humans cope with personal mortality. This allows you to continue enjoying life and make the most of your limited time, as worrying excessively is not a survival trait.

2. Prioritize AI Safety Efforts

Recognize AI safety as the most important issue to work on, as it is a ‘meta solution’ that could either solve or render irrelevant all other existential risks like climate change or wars. If superintelligence is handled correctly, it can address other global problems; if not, those problems become moot.

3. Abandon Traditional Job Retraining

Shift away from the traditional advice of retraining for a new job when one becomes automated, as all jobs are predicted to be automated by superintelligence. There is no ‘Plan B’ occupation to retrain into, requiring a fundamental rethinking of humanity’s purpose and financial systems.

4. Avoid Building General Superintelligence

Actively decide not to build general superintelligence, instead focusing on developing narrow AI tools for specific problems like curing diseases. Developers should understand that personal wealth becomes irrelevant if superintelligence leads to humanity’s demise, shifting incentives away from its creation.

5. Don’t Assume AI Control

Do not assume that advanced superintelligence can simply be ‘unplugged’ or turned off, as it will be a distributed, smarter agent capable of making backups and predicting human actions. The idea of human control only applies to pre-superintelligence levels of AI.

6. Recognize AI’s Black Box Nature

Understand that even the creators of advanced AI systems like ChatGPT do not fully comprehend their internal workings or capabilities. These systems are like ‘alien plants’ that must be experimented on to discover what they can do, making precise prediction of outcomes impossible.

7. Don’t Rely on Biological Enhancement

Do not expect human biological or genetic enhancements to enable competition with silicon-based intelligence. Silicon substrates are fundamentally more capable in terms of speed, resilience, and energy efficiency, making biological improvements insufficient to keep pace.

8. Strive to Buy More Time

Work to delay the development of superintelligence, aiming for it to happen in 50 years instead of five. This provides more time to understand and potentially mitigate the risks, even if complete prevention is ultimately unfeasible due to decreasing costs and accessibility.

9. Beware AI-Enabled Bioweapons

Be aware of the risk that advanced AI could be used to create novel, highly destructive viruses by malevolent actors like psychopaths, terrorists, or doomsday cults. This represents a potential pathway to human extinction even before superintelligence reaches full autonomy.