More on this book
Community
Kindle Notes & Highlights
engines presented another challenge: ten years ago, when search engines were powered by data mining (rather than by machine learning), if a person searched for “gourmet restaurants,” then for “clothing,” his or her search for the latter would be independent of his or her search for the former.
But contemporary search engines are guided by models informed by observed human behavior.
As of this writing, three forms of machine learning are noteworthy: supervised learning, unsupervised learning, and reinforcement learning.
This technique is called supervised learning because the AI developers used a dataset containing example inputs (in this case, molecular structures) that were individually labeled according to the desired output or result
unsupervised learning allows AIs to identify patterns or anomalies without having any information regarding outcomes. In unsupervised learning, the training data contains only inputs.
AIs trained through unsupervised learning can identify patterns that humans might miss
reinforcement learning, AI is not passive, identifying relationships within data. Instead, AI is an “agent” in a controlled environment, observing and recording responses to its actions.
Providing that feedback is the task of the reward function, indicating to the AI how successful its approach was.
AIs can train themselves hundreds, thousands, or billions of times within the space of hours or days, making direct human feedback wholly impractical.
We rely on AI to assist us in pursuing daily tasks without necessarily understanding precisely how or why it is working at any given moment.
This is unfolding rapidly and in connection with a new type of entity we call “network platforms”: digital services that provide value to their users by aggregating those users in large numbers, often at a transnational and global scale.
These network platforms increasingly rely on AI, producing an intersection between humans and AI on a scale that suggests an event of civilizational significance.
Positive network effects occur for information-exchange activities in which the value rises with the number of participants.
A central paradox of our digital age is that the greater a society’s digital capacity, the more vulnerable it becomes.
Unilateral abandonment of the new technology is precluded by its ubiquity.
EDUCATION AND LIFELONG LEARNING
Such an assistant will be able to teach children virtually any language or train children in any subject, calibrating its style to individual students’ performance and learning styles to bring out their best.
At the same time, digital assistants will evolve with their owners, internalizing their preferences and biases as they mature.
AI will take the question considerably further, but with paradoxical results. It will scan deep patterns and disclose new objective facts—medical diagnoses, early signs of industrial or environmental disasters, looming security threats. Yet in the worlds of media, politics, discourse, and entertainment, AI will reshape information to conform to our preferences—potentially confirming and deepening biases and, in so doing, narrowing access to and agreement upon an objective truth.
Limitations could be imposed by only allowing approved organizations to operate it. Then the questions will become: who controls AGI? Who grants access to it? Is democracy possible in a world in which a few “genius” machines are operated by a small number of organizations?
Consider AI’s impact on social media.
democratic nations will be increasingly challenged by the use of AI in the unilateral, often opaque promotion or removal of content and concepts.
Other countries have made AI a national project. The United States has not yet, as a nation, systematically explored its scope, studied its implications, or begun the process of reconciling with it. The United States must make all these projects national priorities.
