Small print.
Wide open.

[Embedded graphic 1 — OCR text]
We believe that Al is one of the most consequential technologies in human
history. If managed well, it could rapidly accelerate scientific innovation,
greatly increase economic growth rates, bring cures for major diseases within
reach, and solve problems that have resisted generations of human effort.

Anthropic aims to realize these benefits while
navigating and preventing the risks that come witha
technological revolution this fast and this powerful.

There are four things that we would like any investor
in Anthropic to understand: the incredible potential of
Al, the risks we must address, Anthropic’s unique
approach to building a company and operating in the
world, and the ways in which our behavior may differ
from that of a purely profit-maximizing company.

Al’s potential

In 2020, four of our founders co-wrote a paper that
documented the scaling laws of AI. The scaling laws
describe how increasing compute, data, model size,
and training environments in combination produces a
rapid increase in Al’s general cognitive capabilities
and, in turn, its economic value. To many at the time,
the idea that machine intelligence would so strongly

depend on scale was surprising, but predictions made
on the basis of the scaling laws have proved true again
and again.’

The phenomenon is somewhat analogous to Moore’s
law for computer chips, under which the computing
power of these chips doubled every 12 to 18 months for
several decades. Moore’s law has been the basis of a
succession of industries—from the personal computer
to the internet—that have revolutionized the modern
economy, but we believe that the impact of an
exponential in machine cognition will be vastly greater
even than that of the exponential in computer chips.

' The idea is closely related to Richard Sutton’s concept of the bitter lesson,
which describes how sheer scale has consistently won out over more specific,
complex methods of training AI models. Note that scale is not the only source
of improvements in model capabilities—algorithmic efficiencies (“compute
multipliers”), which allow us to (for example) achieve higher capabilities with
the same amount of compute, have played a major role. But these efficiencies
are themselves found using the scaling laws: researchers test new ideas and
keep the ones that improve the scaling trend the most.