Artificial intelligence was once a science-fiction topic, but it is fast becoming an integral part of daily life. AI is used for both industrial and work-related purposes, and the aim of these machines and gadgets is to make life easier for people. Scientists have developed a vast array of devices that either find information, complete tasks or improve our living environment. One form of AI that is of particular interest to scientists and inventors is virtual assistants. While this is something that already exists, they want to improve the range of virtual assistants and the functions they can perform. Speech recognition and developing the capacity for AI virtual assistants to talk in just one of the areas on which technology experts are now focusing as they aim to create a seamless talking AI virtual assistant. So. what is such a device and how long will it be before we see a seamless talking AI assistant.?
What is it?
An AI virtual assistant is an artificial intelligence device that performs tasks or services for an individual. This is usually done by verbal command, so speech recognition has played a huge part in the development of this technology. Also known as intelligent virtual assistants, these devices respond to commands and those that use synthesized voices are referred to as talking AI assistants. Users can ask their AI virtual assistant questions, play media, complete tasks such as sending emails and even control their home automation devices. Smartsheet says that this technology is constantly advancing and changes are happening fast. Not only are things changing fast, but the number of people using such devices is increasing at a rapid rate. In 2015, it was believed that 390 million people worldwide used these devices. It is estimated that this figure will have risen to 1.8 billion by 2021.
While AI virtual assistants have many benefits and potential uses, there are also currently some drawbacks. One of these is the problems with speech recognition. If the device does not recognize a word, it chooses the closest word it recognizes in its system. This can lead to the device being unable to complete a task or completing a task incorrectly. Although it is a minimal problem, it is something on which technologists are working on to improve. The other main issue refers to the synthesized speech of AI virtual assistants. The speech is often stilted, and words are mispronounced. Again, this is something on which technologists are working to improve as the aim is to develop seamless talking AI virtual assistants. So, at what stage are the developments and how long will it be before we see a seamless talking AI virtual assistant?
Could it work?
Google has always been one of the leaders in this field and now, they may have made a truly usable voice assistant, says Wired. In May 2019, they previewed speech recognition technology that was extremely responsive and did not have any of the glitches that people have come to expect with voice recognition devices. This could indicate that voice control is the next stage of development and a step closer to a seamless talking AI virtual assistant. The new technology was demonstrated at an annual developer conference that took place in Mountain View. The company was also able to boast that they had significantly reduced the size of their speech recognition software, and this means that they can install it into smartphones. Their developments mean that users will get a quicker and more accurate response from their device. To improve the accuracy of speech recognition, Google used deep learning techniques. It is possible that they will use similar techniques to develop the voice aspect of their gadgets to match the capabilities of their speech recognition technology.
The techniques used to capture the semantic relationships between words and apply this to technology is described further by The red Box. They explain that various techniques have been used, including training AI to accurately predict missing words in a text and improving synthesized speech by getting the AI to study thousands of sentences. This not only extends the vocabulary of the devices but also gives them a greater understanding of the use of words in different contexts. There are already many products on the market that are well on their way to becoming seamless, and these devices have a variety of uses both in the workplace and in personal situations. There are devices that take your calls to screen for scam callers and others that can hold a clear enough conversation that they can book a restaurant on your behalf.
Despite these huge advances, there are still some hurdles to overcome. A point made by The Red Box is that while Ai can follow instructions and can say sentences, they do this only by using either scripted lines or sentences that they create via generated data. However, they do not actually understand the sentences, and this limits them both in terms of speech recognition and seamless speech. It is extremely difficult for those working in robotic sciences to create a device that has a true understanding of language to the same level as humans, and it is this that is preventing the development of seamless talking AI virtual assistants. Therefore, it is difficult to say how long it will be before we see seamless talking AI virtual assistants. Scientists and technologists are constantly working to make improvements in this area and to overcome issues with synthesized speech. In fact, there is a race on between the major players to become the first company to achieve this feat. Companies are applying a vast range of techniques in a bid to improve AI devices’ understanding of semantic language and to eliminate glitches. It is possible that seamless talking AI virtual assistants are something that we could see in the future, but there is not yet a company that has come forward to say they know when they will achieve this.