In the 18th century, the Austrian inventor of automatic machines, Wolfgang von Kempelen, revealed to the public the world's first chess machine. He was a life-sized wax figure dressed in an exotic Turkish outfit, a “Turk”, seated at a wooden box with a chessboard on the top lid. Unfortunately, the principle of the "mechanical Turk" was based on deception - in addition to the mechanisms that set the figure in motion, the real chess player was hidden, who led the games.
The invention of Kempelen had many imitators; at the same time, the principle of the “Invisible chess player” remained unchanged, only the appearance of the machine gun and some technical details changed. Nowadays, the concept of a “mechanical Turk” has been rethought, and a direction called crowdsourcing has grown out of it - transferring a part of production functions to an indefinite circle of people for a small fee.
One of the first successful global crowdsourcing projects was Amazon Mechanical Turk
. Within its framework, Internet users can earn money by performing simple tasks that a person can cope with better than a computer. For example, searching for the most appropriate photos, pattern recognition or editing product descriptions. This approach was called "artificial artificial intelligence" (IRS).
The next step towards integrating human experience into text-writing tools was taken by a group of people developing a Microsoft Word plugin called Soylent
. The word processing interface brings crowd input to helping users solve complex written problems, from preventing mistakes and cutting paragraphs to automating tasks such as finding quotes and correcting grammar.
The project team believes that many workers with basic knowledge of the English language are able to help both novice and experienced writers. These people can perform such tasks that the writer may not cope with or will take a long time to complete, such as scrupulous reading of a document to reduce it later or update the list of addresses to add postal codes to it. They can also provide valuable insight into what is written, pointing out errors that the word processor does not notice, and suggest corrections.
Soylent helps the process of writing texts in Microsoft Word (as the most popular word processor), through the integration of paid labor, based on the principle of crowdsourcing, successfully used on the Amazon Mechanical Turk platform. Ordinary people stand behind Soylent - the basic algorithms use the help of "mechanical Turks" for data processing. It consists of three main components:
- Shortn is a text abbreviation service that allows you to reduce the selected text on average to 80% of the original length without changing the meaning and introducing errors.
- Crowdproof is an improved spelling and grammar service in Microsoft Word that allows you to notice and suggest corrections for errors that the built-in tool misses.
- The Human Macro is an interface for unloading and automating routine word processing tasks such as formatting quotes and finding suitable illustrations.
The famous American science fiction writer, journalist and literary critic Bruce Sterling tried the
new “mechanical Turk” in action and does not hide his admiration. “I wonder if this thing could pass the Turing test?” I see no reason why not, ”he believes.
An interesting discussion about the project Soylent on YCombinator