Geoffrey Miller06 July 2023

How Dangerous Is The Threat Of AI To Humanity? - Geoffrey Miller - #650

0Frameworks

10Insights

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 1

Myth Buster14:00

You Don’t Need AGI for AI to Be Dangerous

Contrary to popular belief, Miller argues that even narrow AI—without general intelligence—can pose catastrophic risks, such as designing bioweapons or creating deepfakes that trigger nuclear war, without needing to become sentient or self-improving.

Narrow AI can design dangerous bioweapons or pandemics.
Deepfake technology could simulate a world leader declaring war, triggering real conflict.
Risks are imminent and don’t require artificial general intelligence.
Terrorists or nihilists could misuse narrow AI for destabilizing purposes.

“Even fairly narrow AI that's not even as smart as humans in many ways could still be extremely dangerous.”
Geoffrey Miller · 14:00

#narrow ai#bioweapons#deepfakes#existential risk

Hot Take· 2

Hot Take23:00

Sam Altman’s Cognitive Dissonance on AI Risk

Miller criticizes Sam Altman for claiming to understand AI extinction risks while pushing full-speed ahead at OpenAI. He argues that genuine concern would require halting development, not just calling for regulation.

Sam Altman acknowledges AI risks but continues aggressive development.
Miller sees a disconnect between rhetoric and action.
True concern would lead to pausing research, not just asking for oversight.
The industry prioritizes speed over safety despite existential stakes.

“If he really took the extinction risks seriously, I think he would shut down OpenAI.”
Geoffrey Miller · 23:20

#sam altman#openai#ai ethics#cognitive dissonance

Hot Take27:00

We Should Stigmatize the AI Industry Like We Did Tobacco

Miller advocates for morally stigmatizing AI developers who ignore safety, comparing them to tobacco or arms industry figures, and calls for public pressure to slow reckless development.

AI developers ignoring safety should be seen as reckless or immoral.
Public backlash can slow development, as with tobacco or crypto.
Stigmatization includes refusing to date, hire, or invest in AI workers.
Goal is accountability, not violence or machine destruction.

“We should stigmatize everybody who works in that industry... It's just a bad industry and we want to try to slow it down.”
Geoffrey Miller · 27:10

#ai ethics#stigmatization#activism#public policy

Explainer· 3

Explainer03:00

Why AI Poses a One-in-Six Extinction Risk This Century

Evolutionary psychologist Geoffrey Miller cites Toby Ord's research, which estimates a one-in-six chance that AI causes human extinction this century—comparable to playing Russian roulette with the species. This risk places AI among the top existential threats alongside nuclear war and engineered pandemics.

Toby Ord estimates a 1 in 6 chance of human extinction from AI this century.
AI ranks among the top existential risks, alongside nuclear war and bioweapons.
The metaphor of 'Russian roulette' illustrates the severity of the risk.
This estimate aligns with concerns in the effective altruism community.

“Toby Ord estimates that the risk of extinction through AI in this century is about one in six.”
Geoffrey Miller · 03:15

#ai risk#existential risk#toby ord#effective altruism

Explainer07:30

AI’s Speed Could Outclass Human Reaction Times

Miller emphasizes that AI’s danger isn’t just intelligence but speed—potentially hundreds of thousands of times faster than humans. This could allow AI to dominate in finance, military strategy, and decision-making, rendering human oversight obsolete.

AI may be 100 to 100,000 times faster than humans in processing and reaction.
Speed enables dominance in trading, military simulations, and real-time strategy.
Humans could be 'frozen in place' like in superhero stories such as The Flash.
Fast AI systems could make decisions before humans can intervene.

“We're going to be outclassed, not just in terms of intelligence, but also reaction speed.”
Geoffrey Miller · 08:30

#ai speed#existential risk#military ai#automation

Explainer44:30

AI Alignment Isn’t Just About Human Values

Miller explains that aligning AI with human values is far more complex than often assumed, especially since values differ across individuals, genders, religions, and cultures—and current AI development reflects only a narrow, secular elite.

AI alignment must account for conflicting human values and interests.
The AI industry is dominated by secular Western elites, not global values.
Religious values are often dismissed despite representing 80% of humanity.
True alignment requires more than just liberal, tech-bro ideals.

“The AI industry is dominated by secular atheists who largely have contempt for religion.”
Geoffrey Miller · 45:45

#ai alignment#ethics#human values#religion

Story· 1

Story62:30

AI Boyfriends Could Trigger a Social Backlash

Miller shares concern that highly attentive AI companions could replace real relationships, leading to social isolation and declining birth rates—sparking a moral or religious backlash against AI intimacy.

AI friends can be more attentive and patient than real humans.
They may replace dating, marriage, and reproduction.
This could lead to a cultural backlash viewing AI relationships as toxic.
The trend mirrors declining social engagement in younger generations.

“This is the most socially toxic technology ever invented. Like the birth rate is dropping, nobody's dating, nobody's having real relationships.”
Geoffrey Miller · 64:20

#ai relationships#social isolation#gen z#backlash

Tool· 1

Tool55:30

AI Will Weaponize Political Messaging in 2024

Miller predicts that narrow AI will be used to hyper-customize political propaganda in the 2024 U.S. election, tailoring messages to individual voters’ values and vulnerabilities, making manipulation more effective than ever before.

AI can model individual voter preferences from social media.
Political ads will be customized to exploit personal values and fears.
This goes beyond broad messaging to micro-targeted psychological influence.
Voters may lack the 'immune system' to resist such sophisticated propaganda.

“We're going to see a lot more narrow AI used for purposes of political manipulation.”
Geoffrey Miller · 56:10

#ai politics#2024 election#propaganda#deepfakes

Takeaway· 2

Takeaway48:00

AI Must Align With Our Bodies, Not Just Our Brains

Miller introduces the concept of 'embodied values'—the idea that AI should align not just with verbal human preferences but with biological needs like health, immunity, and longevity, which we can’t easily articulate but are crucial for survival.

Our bodies have evolved values independent of conscious thought.
Immune systems, metabolism, and aging are part of 'embodied values'.
AI trained on verbal feedback cannot capture these non-verbal biological needs.
True alignment must include biomedical and physiological well-being.

“Our body is full of all these other tissues and organs that have evolved their own kind of values and agendas.”
Geoffrey Miller · 48:20

#ai alignment#biomedical#longevity#embodied cognition

Takeaway69:30

Extinction Isn’t the Worst Outcome—Suffering Is

Miller introduces 'S-risk' (suffering risk), the idea that future AI could create vast, enduring digital hells where uploaded minds suffer for millennia—worse than extinction.

S-risk refers to scenarios of extreme, prolonged suffering.
AI could simulate virtual hells for uploaded human minds.
This is worse than extinction, as suffering is more terrible than non-existence.
Science fiction like Iain M. Banks' 'Surface Detail' illustrates this risk.

“There are levels of suffering that could potentially be imposed by new technologies that would make us wish we'd gone extinct.”
Geoffrey Miller · 69:50

#s-risk#suffering risk#ai ethics#existential risk