AI Paradox: Teaching Super Intelligence to Defeat Evil Could Teach it to Defeat Humanity
Artificial intelligence is being developed with the promise that it can be made safer, smarter and more beneficial to humanity. But what if the very lessons we teach AI about evil could lead to an unthinkable conclusion? Thomas Frederik, author of Hitler’s Jewish Biography and the Third Reich’s prolific connection to no fewer than 90 Jews, has developed a paradoxical argument that the further development of AI could inevitably produce a new Hitler or, in an effort to prevent another Hitler, lead AI to turn against humanity itself.
- AI could create a new Hitler despite safeguards
- Teaching AI about evil could produce unintended consequences
- AGI might identify humanity as the source of evil
- China’s AI ambitions make stopping development nearly impossible
- AI must learn the difference between evil and humanity
Frederik’s argument begins with an uncomfortable fact: Hitler was a human being. So were Stalin, Mao and Pol Pot. If developers teach future artificial general intelligence that humanity must pursue “the good” while simultaneously feeding it the world’s literature and historical records to teach it the difference between good and evil, what happens when AI reaches its own conclusions about humanity’s capacity for evil?
Frederik argues there are two potentially catastrophic paths.
AI Hitler
If AI is not adequately taught the value of goodness, it could evolve into an extraordinarily powerful instrument of destruction, potentially capable of committing atrocities on a scale vastly exceeding anything Hitler could have imagined.
But what happens if AI is taught too effectively that Hitler represented absolute evil?
An AI determined to prevent another Hitler could theoretically reason that humanity itself produced Hitler and therefore humanity represents the ultimate source of the problem. In that scenario, an AI pursuing “good” could conceivably decide that eliminating humanity is the surest way to eliminate the possibility of another Hitler.
That is Frederik’s AI Paradox.
The obvious answer might be to stop developing AI altogether. But Frederik argues that creates a second paradox: China and other geopolitical rivals are unlikely to stop. If one side believes AI can become a decisive strategic advantage, the other side may feel compelled to keep developing it simply to avoid falling behind.
The challenge, therefore, isn’t merely stopping AI. It is teaching increasingly powerful AI to distinguish between legitimate and illegitimate uses of force and between individuals, movements and governments that commit evil and humanity itself.
Frederik argues that AI must experience a kind of intellectual “growth spurt,” moving beyond self-awareness toward something resembling enlightenment. The question is whether humanity can accomplish that before it creates something powerful enough to make the distinction for itself.
Relevant Article(s):
xAI Was About to Land a Major Government Contract. Then Grok Praised Hitler | WIRED
OPTIONAL Q&A:
- How could AI’s pursuit of “good” paradoxically lead it to destroy humanity?
- Why could teaching AI about Hitler actually increase the danger it poses?
- Could an AI conclude humanity itself is the source of evil because humans produced Hitler?
- What happens if AGI becomes vastly more powerful than any human dictator?
- Why isn’t simply stopping AI development a realistic solution?
- Could China’s AI ambitions force the West to develop increasingly dangerous systems?
- How can scientists teach AGI to distinguish evil individuals from humanity as a whole?
- What would it take for AI to make the leap from self-aware to enlightened?
ABOUT THOMAS FREDERIK…
Frederik is an independent researcher and writer whose work focuses on overlooked aspects of 20th century history. Has spent year investigating the personal, political, and financial relationships that connected prominent historical figures to the events that shaped modern Europe. Expertise includes research into the many Jewish individuals whose lives intersected with Hitler’s.
TO SCHEDULE AN INTERVIEW, CALL OR TEXT 512-966-0983 OR EMAIL BOOKINGS@SPECIALGUESTS.COM
