Systems that are able to perform routine tasks instead of humans help support the global economy. Actually, they have been doing this for decades – such systems have been working for the benefit of civilization for a long time. But what about systems that can communicate and interact with the user? If we talk about full-fledged communication, then so far it is not there, there is only one reason – our speech is too complicated.
Everyone who has tried interacting with digital assistants like Alexa and Siri has seen and felt the difference between “talking” with them and having a full-fledged conversation with a person. But progress does not stand still. In the near or not so future, advanced AI interlocutors will definitely appear, because everything is going to this. Under the cut – a discussion of three innovative technologies that stimulate the development of the entire industry.
Conversational AI for processing orders / customer requests
Artificial voice experts have prioritized technologies that can make routine tasks easier, freeing up people to engage in high-impact creative endeavors. An example is communication with customers who order a product or service, processing their orders and requests.
It seems that everything is simple – we load the menu, use the chat bot and that’s it. In fact, there are many factors that complicate things. For example, a system that needs to communicate with customers via conversation requires a near-perfect speech recognition engine that is not affected by car noise, music in the customer’s premises, or any other sounds, including the speech of other people who are near the customer. placing an order. Moreover, the system must recognize the speech of children, adults, people with speech defects, etc.
If you talk on the phone in your apartment, where the music is turned off and there are no extraneous sounds, this is one thing. Many currently existing AI systems can handle such a speech. It is quite another matter to recognize the audio stream from the side of the client, who is on the street or in transport.
The American company Hi auto has overcome these problems, and the system they have developed is capable of operating with 90% accuracy.
Experts suggest that in about 3 years, many restaurants will use an AI voice ordering system. In a few years, this technology will become mainstream – because it will help relieve employees of restaurants and cafes from taking orders.
Conversational AI systems in the clouds for smart machines
The second promising technology that experts highlight is a system that understands the context of a conversation. People usually talk within a certain context, and the same words and phrases can mean different things in different contexts. Understanding context is a natural practice for humans, but not for digital systems. They understand speech literally, without paying attention to the context – the vast majority of digital assistants are simply not capable of this.
Digital assistants are not able to understand even the simplest jokes, let alone double/triple bottom jokes. Actually, even not all people are capable of understanding “complex” humor, let alone cars.
But understanding context is a critical element of a truly effective conversational AI system. Now different companies are working on context-sensitive AI, which, in the course of interaction with the interlocutor, creates models that use additional information, in addition to the personality of the speaker.
A potential area of technology application is chat bots. Ideally, they should collect additional information from different places, including the user’s profile, previous orders, etc. This data can be used to form highly intelligent responses.
Another option is rapid response systems. For example, a person got stuck in an elevator, immediately announced this by voice to the built-in AI system, which instantly notifies the services that are involved in the incident. Here, of course, you can still recall the video with the Scots in the elevator with voice control, but I think this problem is easier to solve than the task of understanding the context of the conversation with a digital system.
Data processing automation
Audio is just one form of unstructured raw data. There are other forms, and all this requires prompt processing, analysis and interpretation.
One example of the use of such technology is the detection of errors in the process of reading a text by a child. One of the largest American educational companies provides a reading aloud service. Children read the text, and the AI system detects errors and, after reading the entire text, shows statistics and a detailed description of the errors.
This is just one of the examples, in fact, there are many more application points for the capabilities of AI systems. In addition, the technologies mentioned above are not the only promising areas. There is also the recognition of emotions, the transformation of speech into text and, conversely, text into speech, and with emotional content.
Now the entire industry of AI voice systems is actively developing, gradually changing various fields of activity and markets in general. Many technologies that are developing now are able to replace a person, saving him from solving boring / routine tasks. This is already happening now, with the passage of time this trend is only increasing.