I think the threat of AGI would be easier to evaluate if it had a firm definition. Right now it just seems like a woo woo threat thrown around by AI promoters. Also, the threat from AI would be based on the viability of the business model, which you touch on a bit. Who is going to pay for its compute and inference costs? Will it be consumers, producers, investors, public markets, tax payers? And which company/companies will own the one AGI that works and how will that survive anti-trust proceedings? Once it’s made what will prevent a deep-seek version being 95% as good and free and destroying the business model?
Just having a vaguely defined fictive computer mega brain running around mad as Norwich is frightening, but the company making it still needs to be solvent I would think and that means someone besides PE has to cover the massive costs of substituting the chips and building new megadatacenters every few years. Where is that money coming from?
Yeah, I think the essay is good but has a blind spot in regards to institutional architecture. The Lindblom mention assumes democracy is mostly periodic elections and therefore thinks that ordinary people have only indirect control over the forces shaping society. But this is wrong, historically, lower case "d" democratic systems worked through thousands of perpetual access points into decision-making, municipalities, locally governed schools and colleges, civic organizations, locally anchored financial institutions, mass-member parties, regional business ecosystems, and several other varying intermediary bodies.
And relatedly, AI does not arrive into a vacuum. Even before deployment, its development is already being shaped by ownership structures, capital allocation systems, legal frameworks, regulatory jurisdictions, educational systems, and organizational incentives. Whether AI ends up serving people or people end up serving AI will institutional landscape into which the technology is embedded and thats also what determines whether things are democratic since lower case "d" democracy is mostly an institutional design and layout matter.
A widely distributed, federated decision-making architecture is likely to produce very different outcomes than one in which a small number of corporate, financial, governmental, or technocratic centers dominate decisions
An interesting essay, but I think it has a blind spot in regards to institutional architecture. The Lindblom mention assumes democracy is mostly periodic elections and therefore thinks that ordinary people have only indirect control over the forces shaping society. But this is wrong, historically, lower case "d" democratic systems worked through thousands of perpetual access points into decision-making, municipalities, locally governed schools and colleges, civic organizations, locally anchored financial institutions, mass-member parties, regional business ecosystems, and several other varying intermediary bodies.
And relatedly, AI does not arrive into a vacuum. Even before deployment, its development is already being shaped by ownership structures, capital allocation systems, legal frameworks, regulatory jurisdictions, educational systems, and organizational incentives. Whether AI ends up serving people or people end up serving AI will institutional landscape into which the technology is embedded and thats also what determines whether things are democratic since lower case "d" democracy is mostly an institutional design and layout matter.
A widely distributed, federated decision-making architecture is likely to produce very different outcomes than one in which a small number of corporate, financial, governmental, or technocratic centers dominate decisions
I don't think we need wait for AI to take on blue collar work to cause chaos, demand would fall off a cliff if white collar workers are rendered obsolete, and blue collar wages would plumet surely, seems like an economic death spiral. Otherwise I'm quite optimistic, even in my work as an accountant I feel it slowly encroaching
I think the threat of AGI would be easier to evaluate if it had a firm definition. Right now it just seems like a woo woo threat thrown around by AI promoters. Also, the threat from AI would be based on the viability of the business model, which you touch on a bit. Who is going to pay for its compute and inference costs? Will it be consumers, producers, investors, public markets, tax payers? And which company/companies will own the one AGI that works and how will that survive anti-trust proceedings? Once it’s made what will prevent a deep-seek version being 95% as good and free and destroying the business model?
Just having a vaguely defined fictive computer mega brain running around mad as Norwich is frightening, but the company making it still needs to be solvent I would think and that means someone besides PE has to cover the massive costs of substituting the chips and building new megadatacenters every few years. Where is that money coming from?
Yeah, I think the essay is good but has a blind spot in regards to institutional architecture. The Lindblom mention assumes democracy is mostly periodic elections and therefore thinks that ordinary people have only indirect control over the forces shaping society. But this is wrong, historically, lower case "d" democratic systems worked through thousands of perpetual access points into decision-making, municipalities, locally governed schools and colleges, civic organizations, locally anchored financial institutions, mass-member parties, regional business ecosystems, and several other varying intermediary bodies.
And relatedly, AI does not arrive into a vacuum. Even before deployment, its development is already being shaped by ownership structures, capital allocation systems, legal frameworks, regulatory jurisdictions, educational systems, and organizational incentives. Whether AI ends up serving people or people end up serving AI will institutional landscape into which the technology is embedded and thats also what determines whether things are democratic since lower case "d" democracy is mostly an institutional design and layout matter.
A widely distributed, federated decision-making architecture is likely to produce very different outcomes than one in which a small number of corporate, financial, governmental, or technocratic centers dominate decisions
An interesting essay, but I think it has a blind spot in regards to institutional architecture. The Lindblom mention assumes democracy is mostly periodic elections and therefore thinks that ordinary people have only indirect control over the forces shaping society. But this is wrong, historically, lower case "d" democratic systems worked through thousands of perpetual access points into decision-making, municipalities, locally governed schools and colleges, civic organizations, locally anchored financial institutions, mass-member parties, regional business ecosystems, and several other varying intermediary bodies.
And relatedly, AI does not arrive into a vacuum. Even before deployment, its development is already being shaped by ownership structures, capital allocation systems, legal frameworks, regulatory jurisdictions, educational systems, and organizational incentives. Whether AI ends up serving people or people end up serving AI will institutional landscape into which the technology is embedded and thats also what determines whether things are democratic since lower case "d" democracy is mostly an institutional design and layout matter.
A widely distributed, federated decision-making architecture is likely to produce very different outcomes than one in which a small number of corporate, financial, governmental, or technocratic centers dominate decisions
I don't think we need wait for AI to take on blue collar work to cause chaos, demand would fall off a cliff if white collar workers are rendered obsolete, and blue collar wages would plumet surely, seems like an economic death spiral. Otherwise I'm quite optimistic, even in my work as an accountant I feel it slowly encroaching