Artificial intelligence and automatic learning machines: what can be protected and how
Artificial intelligence as a mathematical method
Artificial intelligence and automatic learning machines become patentable when they are linked to a specific technological application.
Without it, these computational models and algorithms cannot be protected by patents as they are mere abstract concepts with a mathematical nature.
Is it enough to simply add an electronic device to patent an artificial intelligence?
No. Artificial intelligence should give a technical contribution to the invention and cannot be a later artificial addition. Moreover, the invention entailing artificial intelligence should have a technical nature.
When does artificial intelligence have a technical nature?
Some examples may help elucidating the matter:
- An artificial neural network used in an electronic device monitoring the heartbeat in order to identify irregular beats has a technical nature.
- Classification algorithms ranking digital images, videos, audio and voice signals based on technical features of low resolution and quality, and that are difficult to distinguish, have a technical nature.
- An artificial intelligence controlling an X-ray device or a metal cooling process has a technical nature.
- An artificial intelligence setting the number of passes required to a compaction equipment in order to obtain the desired material density has technical nature.
- An artificial intelligence improving or analyzing a digital audio, image or video, for example by reducing the noise, detecting people in a digital picture, or assessing the quality of a transmitted digital audio signal has a technical nature.
- An artificial intelligence cleaning up background noises in audio files has a technical nature.
- An artificial intelligence identifying and transcribing words has a technical nature.
- An artificial intelligence codifying and decoding, encrypting or decrypting data for effective and reliable transmission and storage has a technical nature.
- An artificial intelligence optimizing a network load distribution has a technical nature.
- An artificial intelligence defining a subject’s energy expenditure by processing data obtained through physiological sensors, deducing the subject’s body temperature thanks to the data gathered with a heat detector in his/her ear has a technical nature.
- An artificial intelligence providing a person’s genotype estimation based on the analysis of DNA samples, and giving a confidence interval of this estimation in order to assess its reliability, has a technical nature.
Can educational methods for artificial intelligence be protected?
It is possible to patent training methods for artificial intelligence as long as they have a technical nature. When a classification method pursues a technical purpose, then the passages to generate a training sample educating artificial intelligence may contribute to the invention’s technical nature, if it supports the fulfillment of such technical purpose.
When is it not possible to get a patent?
An artificial intelligence classifying text documents only for their textual content is not patentable, as it pursues a mere linguistic purpose and not a technical one.
An artificial intelligence classifying records of abstract data or telecommunication networks data with no indication of a technical use of the resulting classification is not patentable because it does not have a technical nature.
Can the results obtained thanks to artificial intelligence can somehow be protected?
Databases or results extracted through the use of some artificial intelligence can be protected by Trade Secret.
What if nothing can be patented?
If artificial intelligence or automatic learning machines are not patentable, it is still possible to protect the research and development in two ways with regard to the specific case at hand: either by making the algorithm secret with a software protection procedure, or by protecting the source code as codified by copyright law.