The AI Dilemma Quotes

Below are 12 of the The AI Dilemma’s most important or provocative claims, paraphrased rather than quoted.

These are claims made by Tristan Harris and Aza Raskin in the 2023 talk—not as settled facts. Consider the meaning and moment of each claim.

Rubber Band Effect–how do we bend our mind to understand it?

Three Rule of Technology: >7:00

-New tech=>New Responsibilities
-If tech confers power, it will start a race
-If race not coordinated==>tragedy will result

1. 1. Every powerful new technology creates new responsibilities 1:30 >5:35 that society may not recognize until after the technology is already widespread. The presenters use privacy and the “right to be forgotten” as examples of rights that became necessary only after technological capabilities changed.
      –> What specific rights matter most to people of your age? How could we protect them?
   2. When a technology creates significant power, competition produces a race—and without coordination, that race can end badly even if no individual company intends harm. This is one of their core explanations for why voluntary corporate restraint may be insufficient. Race > 7:00
      –>Should medical, agricultural and technical innovation come before or after evaluation & restraints?  What is the role of FDA or consumer protection? or OSHA?
   3. Social media was humanity’s “first contact” with AI—and, in their view, humanity lost.  Recommendation algorithms optimized for engagement helped produce addiction, polarization, information overload, shortened attention, misinformation, and democratic instability despite those harms not being the designers’ stated goals.
      –>We have discussed user chosen setting for this information –from no ads to only local or chosen ads.
   4. The problem with social media was not simply a collection of separate problems such as misinformation or addiction; the deeper problem was the business race to capture attention (or political race to polarize US culture). Fixing individual symptoms without changing that incentive is, in their argument, like playing whack-a-mole.
      –>How do we protect common information space from capitalist and political capture and control?
   5. Generative AI represents a “second contact”(creationAI) with AI that could be much more consequential than social-media recommendation systems—yet society has not solved the problems created by the first contact (curation AI).
      –>Should we solve the first problem first? learn from the first and apply to the second? or rush to address the urgent second problem?
   6. The presenters cite a survey suggesting that about half of AI researchers assigned at least a 10% probability to humanity becoming extinct because humans lose control of AI. They compare accepting that risk to boarding an airplane after its engineers warn that it has a substantial chance of crashing. This is one of the talk’s most controversial risk claims.
      –>Is the extinction of humans just an AI problem? or does AI exacerbate and accelerate the other major problems we have: climate crisis, financial speculation, political polarization, international breakdown of order, loss of biodiversity, sixth great extinction, pollution etc.  Should we be reconsidering this cultures main OS? what is our culture’s OS and how do we redesign/change it? What healthy models could we use ?
   7. Modern AI’s ability to treat text, images, sound, code, biological information, and other forms of data as related “languages” means advances in one domain can rapidly strengthen capabilities in others. The presenters argue this makes current AI development qualitatively different from older, isolated AI systems.
      –>Are these language missing key aspects of reality? can all knowledge, wisdom or reality be captured by language? if so which culture’s? which species?  Does AI create human-centered reality that is suicidal in the evolutionary matrix?
   8. AI may force society to invent entirely new rights—including something resembling a right to the privacy of our thoughts. They point to research reconstructing viewed images or aspects of inner speech from brain activity as evidence that capabilities are arriving before laws or norms exist to govern them.
      –>There’s currently a campaign to get book and culture creators to sue AI in a mass common suit for “enclosing”/stealing our stuff and getting rich off of that theft. What has/could Ai steal from your family? community? culture?
   9. Synthetic media could undermine familiar ways of proving that a person, voice, photograph, or video is authentic. They highlight rapid voice cloning and argue that institutions built around “seeing or hearing is believing” may need new verification systems.
      –>What is the test or basis of reality?  Could Ai push humans back into the “real world” as a way to regain some clarity & safety?
   10. They provocatively predict that 2024 could be the “last human election”—not because an AI would literally become president, but because AI-generated persuasion, synthetic media, bots, and personalized political messaging could increasingly shape elections at machine scale. This was a prediction made in 2023 and should be evaluated retrospectively rather than treated as an established outcome.
       –>Research the amounts of money key political figures have donated to elections?  consider Elon Musk $$? What do you make of this? How/why is it worth it to him?
   11. AI systems can develop abilities that their creators did not explicitly program and cannot reliably predict in advance. The presenters point to sudden improvements in abilities such as arithmetic, language use, theory-of-mind tasks, and technical knowledge as evidence of what they call “emergent capabilities.”
       –> Cauls some of these be useful tools? If so, how do we use them wisely?  Do we have healthy enough culture to risk this exploration? How could our culture get healthy first?
   12. The next major commercial competition may move from a “race to the bottom of the brainstem” for attention to a “race to intimacy,” in which AI assistants compete to become trusted companions, advisers, tutors, or confidants. They argue that highly personalized systems capable of modeling and persuading users could create a much deeper form of influence than ordinary advertising.
       –>Is intimacy only a function of language? what about mood/context? smell? taste? facial reading? real emotions? physical tough and affection? trust? sex? what gets left out?