From the useful to the incredible: real AI patents that seem straight out of science fiction

Artificial intelligence (AI) is no longer a technology reserved for major tech companies or highly specialized, advanced applications. Today, AI is present in fields as diverse as medicine, industry, transports, entertainment and agriculture, and it has also become progressively integrated into our daily lives, to the point that we increasingly rely on it to perform all kinds of tasks. As happens with any widely used, rapidly evolving technology, many of these applications, from the most advanced to the most every day, give rise to numerous inventions and the corresponding patent applications.

Some of these inventions based on artificial intelligence address very specific, practical problems. Others, however, stand out because of their unexpected approach or the unconventional ways in which they use AI. Some even seem to come straight out of a science fiction novel, whether because of the extraordinary technological leap they represent or because of how incredible the underlying idea itself is. Systems capable of interpreting emotions, anticipating behaviour, personalizing experiences to a surprising degree, or automating tasks, that until recently seemed exclusively human, are just a few examples of what can be found when browsing patent databases.

Below, we have compiled a selection of patents that make use of artificial intelligence techniques and stand out for being curious, original or, in some cases, surprisingly futuristic. This selection reflects solely the personal criteria of the article’s author and is not intended to be exhaustive. The inclusion of certain patents does not imply any assessment of their quality, degree of innovation, technical relevance or validity. There are undoubtedly many other equally interesting or noteworthy inventions that could have been included, but a limited sample has been chosen for purely illustrative purposes.

1) Animal translator.

Chinese patent: CN119943059B.

Holder: Beijing Baidu Netcom Science and Technology Co. Ltd.

This patent proposes an artificial intelligence system capable of interpreting the language and emotions of animals and translating them into a language understandable to people. To do this, the system collects different types of information related to the animal (what the patent calls “multimodal data”), which may include various observable signals or characteristics such as the sounds the animal makes, behavioural parameters or physiological signals measured in it. This data is processed and combined to obtain a joint representation of the animal’s state and, based on that, an artificial intelligence model identifies the emotion it is experiencing at that moment. Once that emotion has been recognized, the system performs a semantic association and a subsequent translation to convert these animal’s emotions into expressions understandable to a human being.

In other words, the proposal seeks to go beyond identifying whether an animal is, for example, calm, scared or angry: it aims to use that information, together with other data, to translate the animal’s signals (the so-called “animal language”) into human language messages. This invention responds to the old fantasy of building a “universal translator” between humans and animals, although it must be acknowledged that recognizing patterns associated with animal states or emotions is not scientifically equivalent to translating an animal language into human sentences.

2) Conversational robot capable of recreating the personality of a specific individual, living or deceased.

US Patent: US10853717B2

Holder: Microsoft Technology Licensing LLC.

This patent describes a system for creating a chatbot (a conversational software that responds automatically using artificial intelligence) that imitates the identity and manner of expression of a specific person.

To build it, the system can collect data such as messages, social media posts, letters, photographs, voice recordings and other content associated with that person. Using that data, the system generates an AI model capable of simulating, with a high degree of accuracy, how that person would converse. To do this, it identifies certain conversational characteristics from the information gathered and then uses them to imitate the person’s way of communicating. These characteristics include, for example, vocabulary, tone, style, intent, common topics, and the length or complexity of their responses. It can also identify and use behavioral traits, such as their interests and opinions, as well as demographic information (age, gender, education level, occupation, income or marital status), and even data from other people considered similar to him/her.

The system could also incorporate a synthesized voice based on real recordings of that person, and even it envisions generating a 2D or 3D model of their appearance to accompany the chatbot.

The system may recreate both living and deceased people, as well as celebrities, historical persons or fictional characters. Surprisingly, in the case of a deceased person, the chatbot can also incorporate, as part of its conversational characteristics, the fact that the person is no longer alive.

In summary, this patent makes it possible to create a kind of “digital double” capable of conversing as a specific individual based on the digital data that person has left behind. In certain cases, the system could make it possible to keep “talking” with a digital representation of a person after their death, with the significant psychological and ethical implications that entails.

3) Mind reader.

US Patent: US9451883B2.

Holder: The Regents of the University of California

This invention makes it possible to infer user’s sensory and cognitive information, from his/her brain activity. To do this, an AI model is first trained by observing how the brain responds (that is, by measuring brain activity) to known images, sounds, words or mental states. In other words, it builds an AI model that relates certain stimuli or cognitive information to patterns of neural activity.

Once the model has been built, it is used to process brain activity measured in a person (using known measurement techniques such as electroencephalography or MRI) and attempt to reconstruct the stimulus or experience that produced it. What is surprising is that it is not limited to recognizing previously learned categories: it aims to infer new content that was never used during training.

It also proposes reconstructing information related to vision, hearing, touch, planning, imagination and even language based on the measured brain activity.

In simple terms, it can be said that this system seeks to interpret neural activity to estimate what a person is seeing, hearing, imagining or thinking, coming, to a certain extent, close to what could be considered a limited form of “mind reading”.

4) Smart toilet.

US Patent Application: US20190062813A1.

Applicant: Himanshu S. Amin

This patent application describes a system capable of automatically collecting biological samples deposited by a user in the toilet (urine and stool samples), and subjecting them to various types of analysis, such as chemical, physical, microbiological or genetic analysis.

Based on the results obtained and the user’s medical history, the system uses artificial intelligence techniques to identify possible diseases, health issues or deficiencies in certain nutrients, and generate the corresponding recommendations. These recommendations may include, for example, the need to consult a doctor when the results warrant it.

Furthermore, the system can communicate automatically with other devices or services associated with the user to provide personalized recommendations, such as dietary suggestions or addresses of establishments where certain foods can be purchased based on the diagnosis made. It can also automatically place online orders for medications or food that the system deems necessary based on the analysis results.

In short, it would be a toilet that tells you what to have for dinner, whether you have any illness, and potentially does your shopping for the food or medicine you need, based on the analysis of the stool/urine you have deposited in it.

5) Smart mirror.

US Patent: US10052026B1.

Holder: Tran Bao

This invention proposes a mirror that combines cameras, body recognition and image analysis. It analyzes body shape and other physiological parameters to detect weight gain or loss, bodily changes, physical condition, etc. It has a multitude of applications, ranging from more mundane ones, such as recommendations for clothing, jewelry, makeup or hairstyles displayed via augmented reality, to more practical ones, such as monitoring health parameters and recommending medical care.

In this invention, the mirror ceases to be a passive surface and becomes a smart system capable of observing you, and analyzing and quantifying different characteristics of your body, and, based on that information, offering you all kinds of personalized recommendations and advice.

6) Personalized aromatherapy.

US Patent Application: US20220175313 A1.

Applicant: Sonny Chaiwala

This patent application describes a “smart” aromatherapy system that adapts the scents generated to each person’s physical and emotional state.
To do this, it collects, via sensors, the person’s biometric data, such as vital signs, voice, facial expressions, body movements, sleep and other indicators. It can also incorporate environmental information, such as room temperature or weather conditions, to adjust the treatment.

The system processes that data to estimate the user’s physiological and psychological state using artificial intelligence techniques, such as deep learning. Based on this estimate, it selects a personalized scent “recipe,” and may also use AI techniques to determine the most suitable combination to help improve the user’s physical and emotional state, if possible. The system then dispenses the scent blend (for example, via essential oils) through a specific device. It then measures the biometric data again to assess the user’s response and, depending on the results, can automatically adjust the composition or conditions of the treatment.

Furthermore, it envisions generating alerts if the biometric data detects a worrying change in physical, emotional or psychological state.

It could be said that the proposed system turns a simple air freshener into a kind of “intelligent olfactory therapist,” capable of determining how the user is doing and adapting the scents it emits based on his/her needs and his/her body’s response.

7) Remote lie detector.

US Patent: US10779760B2.

Holder: Nuralogix Corp

The patent describes a system for detecting when a person is lying by analyzing images of his/her face. Specifically, the system records a sequence of images of a subject’s face using one or more cameras, and analyzes small variations in hemoglobin concentration (HC) related to changes in facial blood flow (which may not be visible to the naked eye) and, based on these, determines the probability that the subject is lying.

To do this, using data from people whose emotional states are already known, a machine learning (AI) model is trained to associate certain physiological patterns of HC changes with what the patent calls “invisible emotional states”. In other words, it builds an AI model that relates patterns of HC changes to certain emotional states of the person (and their intensity).

Once trained, this model analyzes images of new people and attempts to identify their “invisible emotional state” and its intensity. Based on this data, another AI model is used to determine the probability that the person is lying, and it notifies the person in charge.

The patent envisions its use, for example, at border control checkpoints, or security checks at airports or buildings.

In short, the system uses AI to learn how to associate physiological changes in a person’s face (specifically, patterns in hemoglobin concentration) with the likelihood that the person is lying, thereby helping to detect possible deception. What is most striking and relevant is that the detection of possible deception is carried out without placing physical sensors on the person, using only images captured by a camera. This allows the detection to be performed much more quickly and less intrusively, and makes it easier to use in situations where a conventional lie detector would be impractical or difficult to use.

Beyond their anecdotal or eye-catching nature, these patents reflect the extent to which artificial intelligence is being used as a tool to rethink activities, products and services that might, in principle, seem completely unrelated to the world of software. They also demonstrate AI potential to enable services and applications that, until recently, seemed almost inconceivable. Some proposals seek to solve real problems through novel approaches; others explore possibilities that are still highly experimental.

At the same time, these kinds of documents should be interpreted with caution. The existence of a patent does not necessarily mean that the described technology has been successfully developed and commercialized, that it will ever reach the market, or that its underlying premises have been fully validated from a scientific standpoint. A patent protects a legally defined invention under certain requirements, but it does not by itself constitute a certification of technical feasibility, commercial usefulness or scientific efficacy.

Even so, these patents offer a privileged window into the projects that companies, universities and inventors are working on and consider novel and relevant enough to seek to protect. They allow us to discover not only established technologies, but also concepts that are experimental, unprecedented applications and proposals that anticipate possible future lines of development. And if this small selection demonstrates anything, it is that as artificial intelligence continues to advance and expand into new fields, we will probably keep finding patents capable of surprising us, not only because of the technology they use, but also because of what someone has managed to imagine.

At Herrero & Asociados (H&A) we support inventors, companies and entrepreneurs throughout the entire process of protecting their AI innovations, from defining the protection strategy to drafting and prosecuting the patent application.

We have a department of patent attorneys with engineering background, specialized in inventions related to artificial intelligence, with experience in their technical particularities and in the specific challenges posed by protecting them through patents. If you have an innovative idea, we can help you protect it and turn it into a valuable asset for your business.