What is Claude? Anthropic doesn’t know either

Gideon Lewis-Kraus at The New Yorker goes in depth with researchers at Anthropic attempting to understand how Claude works and finds AI is even more weird and confusing than we think:

The most candid A.I. researchers will own up to the fact that we are doing this because we can. As [Brown computer scientist Ellie Pavlick] wrote, the field originated with the aspiration “to understand intelligence by building it, and to build intelligence by understanding it.” She continued, “What has long made the AI project so special is that it is born out of curiosity and fascination, not technological necessity or practicality. It is, in that way, as much an artistic pursuit as it is a scientific one.” The systems we have created—with the significant proviso that they may regard us with terminal indifference—should inspire not only enthusiasm or despair but also simple awe.

In the eighteenth century, James Watt perfected the steam engine: a special box of fire that turned archaic fern sludge into factories, railroads, and skyscrapers. The Industrial Revolution happened without any theoretical knowledge of the physical principles that drove it. It took more than a century for us to piece together the laws of thermodynamics. This scientific advance led to such debatably beneficial things as the smartphone. But it also helped us explain why time flows forward, galaxies exist, and our universal fate is heat death.

Now we have a special box of electricity that turns Reddit comments and old toaster manuals into cogent conversations about Shakespeare and molecular biology. The sheer competence of language models has already revamped the human quest for self-knowledge.

Askell describes AI is “neither robot nor human but actually something new.” Based on my use of AI and learning about it, that’s a good way to describe it.

Has AI achieved intelligence? Define “intelligence.” In the 2019 book “Ways of Being: Animals, Plants, Machines: The Search for a Planetary Intellience,” James Bridle argues that our definition of intelligence — which limits the phenomenon to humans and maybe some higher animals — is too narrow. Software is intelligent, as are analog computers, as are bonobos, jackdaws, bees and trees. Bridle is not arguing here that plants have “hidden lives,” like 1970s pseudoscientists argued, but that these machines, birds, animals and plants should be considered intelligent just on the basis of their observed behavior.

I’m skeptical that AI has achieved anything resembling human intelligence. It is not alive. But I’m even more skeptical of people who dismiss AI as just a fancy autocomplete.

Pair the New Yorker article with this essay by Matt Shumer, an AI entrepreneur and investor: Something big is happening.. Shumer says that within a year or two, AI will be better than humans at any job that’s now done at a screen. He compares the present moment to February 2020, the weeks before the pandemic hit, when nearly everyone went about their normal lives but a few people knew that the world was about to profoundly change.