AI Learning – An Exciting Field

July 9, 2018

A few decades ago something that was pure science fiction and originally used in chess software at best, has now become reality in many fields: computer programs that emulate human thought patterns, make decisions, draw conclusions and learn.

Artificial intelligence is becoming increasingly important in daily life: almost everyone has heard of speech recognition like “Siri” and others, several companies are working on self-driving cars, and the subject is becoming more and more interesting and important for clickworker and our customers too.
The always improving imitation of human intelligence and human decisions by machines and algorithms has made rapid advancements especially in the last few years, and the potential here is surely not even remotely exhausted yet. For just this reason we expect exciting, new application areas and jobs especially in the field of “AI learning.”

What Exactly is “AI Learning”?

The abbreviation AI stands for artificial intelligence. “AI Learning” then is the training and improvement of this artificial intelligence. For a computer program, or its algorithm, to be able to imitate a human way of thinking, it needs data to draw on and with whose help it can estimate and evaluate probabilities – a whole lot of AI training data. Speech recognition software for example should reliably recognize a command even if the speaker is currently congested or doesn’t have perfect command of the language. A program for analyzing handwriting should work dependably even if the source text is not in perfect calligraphy or when someone is writing with a bandaged hand. And for self-driving cars it would be desirable for them to reliably recognize a one-way street or stop sign, so that the car reacts accordingly and brakes automatically.

Artificial Intelligence
(Image by user geralt)

The same principal applies for all these systems: The more data a program has to draw on for its decision making, the more likely it is that the correct choice will be made. And this is where clickworkers can come into play. In the learning phase for every AI, raw data must be generated and pre-sorted, decisions checked and evaluated. And as of now only a human can do that reliably. Our large and extensive community has a special advantage precisely here: no matter which languages, countries or technical equipment (i.e. certain smartphones) the customer has in mind, in most cases we have the appropriate clickworkers.

Project Examples of AI Learning on clickworker

In the past particularly developers of speech recognition software have relied on the combined forces of our Clickworkers multiple times, often in several stages too. For example, first commands for a certain action were collected (“How would you command a voice controlled radio to switch to an oldies station?”) and afterwards for the second part of the project the collected commands were spoken and recorded by Clickworkers. With the help of this data the AI learns to reliably recognize and execute instructions.

Take a look at our case study: Speech recognition training data for software development

In the field of text recognition too we have already successfully completed various projects. Recently for example our Clickworkers verified whether the summaries a program automatically created from longer texts retained the meaning of the original text.

Video recognition is also an in demand field: whether faces, gestures, moods or objects, we have successfully implemented the most varied customer requirements and had corresponding video recordings / video datasets made by our Clickworkers.

AI learning remains an exciting field with a huge amount of potential in the near future. And clickworker will be there to help!

Dieser Artikel wurde am 09.July 2018 von Daniel Marz geschrieben.


Daniel Marz - Die - Your crowdsourcing Platform