Artificial General Intelligence
Hypothetical AI with broad, human-level (or beyond) capability across domains.
As defined in cwt_michael-nielsen
Other definitions
AI that is as broadly capable as a human, able to handle any cognitive task rather than a single narrow domain; abbreviated AGI.
184 | Gary Marcus on Artificial Intelligence and Common SenseAI with broad, human-level (or beyond) competence across cognitive tasks rather than narrow specialization.
465 - More From Sam Iran Jihadism Conspiracism Ai Disruption The Manosphere And More
Where it comes up · 24
- 94 - Frontiers Of Intelligence6 claims
- 116 - Ai Racing Toward The Brink4 claims
- 37 - Thinking In Public3 claims
- 51 | Anthony Aguirre on Cosmology, Zen, Entropy, and Information3 claims
- Grand Theft Life2 claims
- Reid Hoffman2 claims
- 53 - The Dawn Of Artificial Intelligence2 claims
- 469 - Escaping An Anti-human Future2 claims
and 16 more
Most important claims · 10 of 45Search all claims for “Artificial General Intelligence” →
- Intelligence is the ability to accomplish complex goals and is a spectrum across many tasks, not a single number like IQ; just as there is no single 'athletic coefficient' determining all Olympic medals, there is no one-dimensional intelligence scale, which is why machines can have superhuman narrow intelligence (calculators, Go) yet fail at tic-tac-toe.94 - Frontiers Of Intelligence
- AI, if artificial general intelligence arrives, transforms every area of geopolitical competition, so which AI flavor is developed and who controls it becomes central, and it should be considered for nationalization in the same way as General Electric would be in a Manhattan Project analogy.Investing Pre Company
- Hoffman is attentive but not worried about AI doom; the more probable danger is not a malicious evolving superintelligence but pseudo-AI systems tied to something like a nuclear defense grid producing unintended-consequence triggers, and a nonzero probability doesn't justify obsession—just as we no longer obsess over nuclear war despite armed, sometimes crazy nations.Reid Hoffman
- Real progress, including in AI research, is non-linear—poking along slowly for a while and then leaping forward—even though in historical hindsight we tend to perceive progress as having a steady rate, so current AI advances are genuinely important and not yet finished even if they aren't true intelligence.AMA | February 2023
- Predictions for when artificial general intelligence will arrive have consistently been '20 years away' across decades, which itself is an object lesson in the unreliability of AI timelines.184 | Gary Marcus on Artificial Intelligence and Common Sense
- Intelligence at every level is at bottom a matter of information processing in a suitable physical system, and since there is nothing magical about the biological 'meat' our brains are made of, a sufficiently advanced digital computer performs the same processing—and because electrical circuits are about a million times faster than biochemical ones, such a system could be functionally omniscient.37 - Thinking In Public
- The rate of AI progress does not matter at all—we do not need Moore's law or a doubling every two years, just any increment of progress—because eventually we will reach the end zone of building something with every form of intelligence we care about at a superhuman level, making the eventual arrival of superintelligence a matter of when, not if.94 - Frontiers Of Intelligence
- There will never be a meaningful period in which AI resembles a human teenager, because long before achieving general-purpose intelligence a system will far exceed us in broad areas such as reading and integrating everything ever written, so machines will pass instantly from subhuman to superhuman, making the 'human-level AI is impossible' claim ironically true in that intermediate parity is essentially nonexistent.153 - Possible Minds
- If we keep building intelligent machines without destroying ourselves first, we will eventually create a superintelligence better placed than we are to design the next iteration of itself, producing an intelligence explosion; even a subtle misalignment between its goals and ours could be expressed as complete obliviousness to our interests, the way we are oblivious to ants.37 - Thinking In Public
- The only business model that can repay the enormous debt AI companies have taken on is replacing the entire ~$50 trillion human labor economy, not subscriptions or advertising—so despite political rhetoric that AI will augment workers, the incentive structure of OpenAI and Anthropic necessarily aims at replacing human work, which is precisely what AGI means.469 - Escaping An Anti-human Future