Breadth v. Depth - AI epistemology favors utopianism

Julius Olavarria 🔸

Comments 2

Sorted by

New & upvoted

Executive summary: The author argues that AI-enabled access to broad knowledge is likely to improve ethical decision-making and cause prioritization, despite concerns about reduced depth of understanding.

Key points:

AI summaries are extending a trend toward greater breadth of knowledge.
The author argues that concerns about shallow understanding may be overstated and that people will adapt to breadth-focused learning.
Broad knowledge is especially valuable for ethics and cause prioritization because it reduces ignorance about important problems.
AI could improve decision-making by synthesizing large amounts of information, provided it is truthful and well-aligned.
Preserving incentives for primary knowledge producers remains essential.
The author suggests that wider access to human knowledge could move society closer to a more rational and desirable future.
The author argues that, for cause prioritization, broad understanding of many issues may be more valuable than deep expertise in a few.

This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.

juneatjunedotkim

Humans lie, and LLMs lie, but less often. Also, most people's version of utopia has more to do with doing less mundane stuff than has to do with knowing more stuff

Comments

Curated and popular this week

What would an animal-aligned AI be aligned to?

Aidan Kankyoku, Anima International·2w ago·Curated 6d ago·15m read

This is a crosspost from the new Animal Welfare Alignment Newsletter by Anima International. You can subscribe on Substack if you are interested in following these efforts. Audio reading also available on Substack. The goals of this post are to: 1. Raise a question I see as crucially important to the goal of aligning AI to animal welfare...

137

Let's taboo the V-word

lincolnq·3d ago·8m read

“How long have you been v*g*n?” This is one of the most common icebreakers at animal protection events. It’s a baseline assumption, and it mostly holds true: if you’re out advocating for animals not to be tortured or abused, realistically these days you are v**n, or close. And it makes for good conversation. It seems fairly safe to assume when you meet strangers. But this assumption is hurting the movement in a way which we don’t always notice: someone new comes into the sp...

Spiro: an update 2.5 years on and a fundraising ask for expansion

Habiba Banu·12h ago·6m read

Summary Back in November 2023 I posted here to launch Spiro and raise our first $198k. Two and a half years later this is an update and a fundraiser for the next step. The short version: we've now reached over-5,900 people with TB preventive medicine, including over 3,000 children under five years old. Our early results have held up well an...

Recent opportunities to take action

These claims are probably novelty-averse

We hear warnings of AI brain fry, which come as a result of focusing on too many AI summaries at once, often across different chatbots and several conversations.^[2] While this is not exactly a scientific term, I understand that there can be serious issues using AI this way. I am also considering the drastic importance of warnings to technology, and the general trend and importance of always having counterforces to new technology (e.g. leaded gasoline).

However, 'breadth engagements' like these are likely the future of education, research, and action-deliberation. This is not to say it is a good thing but it probably will be something that our brains adapt to over time: Which I think is more likely to be a good thing primarily because...

Epistemology and ethics rely on breadth

Ethics runs best when what we know = all possible and available human knowledge = all available paths of action. I argue that knowing more about all of the genocides in the world makes you a better ethical agent. Only knowing about one genocide will focus all 'ethical attention' (attention of an agent that would lead to moral decisions such as donating to cause X) to one area simply out of causal ignorance.

Often the best EA donations are from agents who know as much as possible about triage / cause prioritization.

Of course, a full understanding of causality is impossible. AI summaries - as they improve - will be a huge weapon to uncovering hidden causes so that we, as agents, can make better decisions. This works only if alignment is sound and if AIs report the full truth ^[3]. Humans might adapt to read carefully and interpret when an AI response is filler vs. when it is giving useful, concrete information.

A possible critique is that cognitive offloading is real and is probably making us stupider. But I would argue that what we define as "stupid" is vague, and I would also appeal to the Extended Mind Thesis as a potential counterargument: The most important thing for a human is doing good, and to achieve this, agents must have access to all possible human knowledge and it must be synthesized in a way that inspires action, to which AI summaries are our most powerful tools.

And this all requires protecting those who contribute primary sources to the knowledge base have incentives to do so. This might be the most important social invention of our time: How can we protect journalists or researchers 'on the ground' so that AI will continue to give agents the best causal summaries available for action?

Also, what I mean as action might be understood as broadly "how to live well." In general, though, living well will require at least knowing all the facts possible for the agent, to which AI summaries might evolve to be our best tool.

Utopia

If the principal prerequisite to becoming a perfect agent = knowing everything, then we can get close to utopia (or eutopia) this way. ^[4] Because we can never be perfect, we can hope to get close if AI disseminates the entirety of the human knowledge base to agents that aspire to be rational. I see eutopia only as a continuation of all the good things the internet gave us: Universal access to a complete compilation of human knowledge, but in a better way.

Primary critique

A shallow understanding of a lot of things is worse than a deep and detailed understanding of a few things. I would argue that this doesn't apply to cause prioritization, and if cause prio is the most important moral decision(s) we make, a shallow understanding of a lot of really morally important things equates to better decisions.

Please leave comments and challenge assumptions where possible. For example, "These claims are probably novelty-averse" might be a big assumption.