There Is No AI Really (It’s Just People), with Jaron Lanier

Lanier argues that the prevailing ideology of AI as an ethereal, independent creature is misleading. Instead, AI, particularly large language models, should be understood as a new form of collaboration, combining the work of countless people into a single, vast 'document' similar to Wikipedia. This perspective, while technically equivalent, opens up new avenues for ethical development and problem-solving by acknowledging the human source of all data.

Lanier states, 'You can think of large language model AI as a whole bunch of people whose work was combined into this single document... fundamentally it's the same kind of beast.' He also notes, 'AI is made of people. It's made of data from people.'

Despite massive investment (hundreds of billions of dollars), VR has not achieved widespread adoption because companies force it into existing business models (e.g., big iPad, social network, game) instead of allowing it to be 'VR.' A critical failure point is the lack of diverse development teams, particularly the exclusion of women and non-white individuals, leading to products that cause nausea for significant user populations and fail to serve broad needs (e.g., no decent 3D design tools).

Lanier recounts a story where a major tech company's VR head claimed to have 'solved the nausea problem' but had not tested on women or Asian women, believing it was 'the same for everybody.' The first review from an 'Asian female got sick.' He states, 'Nobody's let VR be VR.'

The internet became 'mean' and addictive because the dominant business model in Silicon Valley is 'influence generation.' Companies gain the ability to influence large numbers of people, and others pay to access this influence (or pay 'blackmail money' not to be excluded). This model drives algorithms to constantly activate the 'fast brain' (fight-or-flight response), leading to users becoming petty, vain, confrontational, nervous, and never satisfied, exemplified by the convergence of personalities like Elon Musk and Donald Trump.

Lanier explains, 'The reason we went into the mean direction is that the only business model allowed in Silicon Valley is influence generation.' He adds, 'What it tends to do is it tends to keep that fast brain stuff constantly activated.'

Current privacy frameworks like GDPR are insufficient because 'privacy' is hard to define and often focuses on controlling information flow, which is impractical for users. True privacy should mean 'freedom from manipulation' and specifically involve a 'prohibition on software that interacts with a human that contains any predictive function about that human.' This would ban targeted, manipulative advertising and algorithms that create a 'model of you' in a feedback loop, ensuring people are 'allowed to be responsible for their own behavior.'

Lanier states, 'What it really needs to be is there has to be a prohibition on software that interacts with a human that contains any predictive function about that human.' He clarifies, 'It would prohibit is advertising that has a model of you that predicts your behavior in a in a feedback loop.'