What did Ada Lovelace write about the Analytical Engine?

Ada Lovelace's most famous work started as a translation job. In 1842 a description of Charles Babbage's planned Analytical Engine appeared in French, written by an Italian engineer named Luigi Menabrea. Lovelace translated it into English and then kept going. Her additions, labeled Notes A to G, ended up about three times as long as the article itself, and they contain some of the clearest thinking about computers written in the nineteenth century.

The result was published in 1843 in a London journal called Taylor's Scientific Memoirs, signed only with her initials, A.A.L. It is still the most detailed account of Babbage's engine from his own lifetime.

Who she was and how the notes came about

Augusta Ada King, Countess of Lovelace, was born Ada Byron on 10 December 1815, the daughter of the poet Lord Byron. Her mother made sure she was taught mathematics and science from an early age, and her tutors included the scientific writer Mary Somerville. It was Somerville who brought her to one of Babbage's Saturday evening gatherings on 5 June 1833, when Ada was seventeen. Soon after, Babbage showed her the working section of his Difference Engine, and she became fascinated with his machines.

In 1840 Babbage presented the Analytical Engine to scientists in Turin. Menabrea wrote up the talks, and his French article was printed in Geneva in October 1842. Babbage's friend, the scientist Charles Wheatstone, asked Lovelace to translate it. She worked on it over about nine months in 1842 and 1843, and the translation with her notes appeared in the September 1843 issue.

What the notes say

The notes range from engineering detail to philosophy. Four ideas stand out.

The engine is a general machine. Lovelace stressed that the Difference Engine could carry out one narrow set of operations, while the Analytical Engine could carry out whatever operations its cards described. She used a lovely comparison: the engine "weaves algebraical patterns just as the Jacquard-loom weaves flowers and leaves."

It can work on more than numbers. She argued that the engine acts on anything whose relationships can be written in its notation. Her example was music: if the rules of harmony and composition could be expressed that way, the engine might compose elaborate pieces of any length. She called the general subject "the science of operations," a study of operations in themselves, separate from what they are applied to.

A complete worked program. Note G sets out, step by step, how the engine could calculate Bernoulli numbers, a sequence that appears often in higher mathematics. It lists each operation, which columns of the machine it uses, and where each result goes. The engine was never built, so the method was never run in her lifetime. Babbage had written earlier programs that stayed unpublished, so historians debate whether "first programmer" is the right title; what is clear is that Note G is widely described as the earliest computer program ever published.

A limit. In Note G she also wrote that the engine "has no pretensions whatever to originate anything." In her view the machine carries out the instructions people know how to give it. It can follow a line of reasoning, but it cannot foresee new truths on its own.

A worked example: reading her ideas in 2019

Each of her main points maps onto something we can test today.

Why it matters

Lovelace died young, on 27 November 1852, at the age of thirty-six. For a century her notes were known mainly to specialists. In 1953 they were reprinted in a book about digital computers, B. V. Bowden's Faster than Thought, and a wider audience discovered how much she had seen. Since then she has become a symbol of the early history of computing; a programming language created for the U.S. Department of Defense was named Ada in her honor.

Her lasting contribution is the separation of the machine from its meaning. Babbage mostly thought about calculation. Lovelace saw that a machine which follows rules on symbols could, in principle, handle any domain that can be described with rules. That idea is one of the starting points for the whole field of artificial intelligence.

What is "Lady Lovelace's objection," and why does it still matter?

In 1950 the British mathematician Alan Turing published a paper asking whether machines can think. He listed objections people might raise, and one of them he named "Lady Lovelace's Objection," quoting her claim that the engine cannot originate anything and simply carries out the orders people know how to give it.

Turing gave two answers. First, he suggested that Lovelace was reporting what the evidence available to her showed, not proving that no machine could ever learn. Second, he took on a popular version of her point: the belief that a machine can never surprise us. Machines surprised him all the time, he wrote, because no one can work out in advance every consequence of a set of instructions. He planned to return to the question with the idea of machines that learn.

That is exactly where the debate sits in 2019. Systems that learn from data, like the one that taught itself to play Go, make moves their programmers did not write by hand and could not predict. Yet they still only do what their training and code allow. Whether that counts as originating something is the question Lovelace raised in 1843, and it is still a good one to ask about any new AI system you meet.