AI superpower: where we are now
Human + AI
Several years ago within AMAI we started developing our own Voice AI engine because we had seen how few solutions there were in the market and how many problems could be solved with voice technologies — since then the necessity has even increased, and people began to have the understanding of practical benefits of AI-products: for example, in September two thousand nineteen Sylvain Duranton of Boston Consulting Group within his TEDx presentation noted that positive cases of AI usage happen with those companies who implement eighty percent of their solutions to help a person but not to substitute their job. On the contrary, the companies who were trying to automize processes altogether bump into “algocracy”…
Algocracy?
According to Sylvain, if AI has to make one hundred percent decisions, it soon starts to behave oddly, naughtily and demand new rules to process the simplest situations which are unknown to it. Imagine, you want to hire people to your company and start communicating with AI:
— I have five cool Python-developers and I want to hire three more like them. Find them.
— This guy seems to suit.
— No, no. Python is a programming language.
— Oh, understood. Then, this one.
— Better. But it’s more important for him to be able to write well on Python. Age is irrelevant.
— Then, this one will definitely suit.
— No, not that much: he is still at school. And why did you write to this guy that he didn’t suit?
— People from this town don’t usually have a good experience of IT-products development.
— Yes, that’s because this town has just 30 thousand people — he might be the only developer in this town.
Like in the case of any bad process automation, expenses on its service outreach the increase in efficacy. Nonetheless, AI technologies market is not only growing but is also being actively implemented into vital spheres: medicine, industry, finance, fully automated cars, fully automated cows (thank you,
).
Let’s look at numbers:
- The market size of AI-technologies (hardware, software, services) is now estimated at $39+ billion.
- Annual AI-technologies market growth is about forty two percent (CAGR). Compare it with eleven and a half percent annual market growth of SaaS-solutions and fifteen percent of cloud solutions.
- Gross value add (GVA) will exceed one trillion dollars in the next аifteen years (biggest profit from AI will be to medicine, car manufacturing, retail and finance). Voice AI will have about two hundred billion by our estimate.
Human + AI
AI starts working when it is implemented in a vital part of a process but not tries to cover the whole process. In the case of hiring people to a team, one HR company has effectively automized the recruiting process not by eliminating candidates but by their rating, having acquired the most labour-consuming task in a recruiting process: “Who to write first?”. And leaving the responsibility to make a decision to a recruiter. Here the dialogue with AI looks different:
— I would recommend looking at this guy first, but if he doesn’t suit, I have a lot more.
— Yes, thank you. We have also received updates on the vacancy, can you adjust the search?
— No problem.
There are a lot of examples of positive usage of AI+Human — especially in the spheres where the responsibility for the result is high:
- Computer vision. In the objects of increased safety an operator controls the activity on more than 10 screens. However, if we delegate the responsibility to observe to AI, we will constantly face false positives. In reality, computer vision only informs the operator about an activity in a definite section and an alleged problem.
- Voice-assistant in an online shop. Experience shows that people become really irritated when talking to a robot for too long and they ask for a human. But the successful example of using a voice assistant is the experience of those companies who use this assistant to connect to a required operator and to define the customer’s name by their voice, so that an operator could receive their purchase history immediately.
- Microsurgery. Bioengineers of Boston Children’s Hospital used microrobots which helped to conduct complicated operations: to do this they were taught to define their place in a heart and to find leakages, and after that they transferred the information to a surgeon who didn’t have to waste tiring hours on this search.
Voice AI
We would like to talk about Voice AI technologies in particular on two accounts: in AI context this is one of the top directions; we are developing Voice AI products ourselves and can see the growing need in them.
Voice AI includes both speech recognition and text-to-speech technologies which allow AI reproduce maximally natural human speech based on a wide range input moods — you can understand how it works in our demo.
Such technologies are becoming an integral part of client service in banks and online shops: eighty two percent of Voice AI market is occupied by these spheres; fifteen percent — by car industry. Among the directions that we are developing now we would like to note the following:
- Call-centres. It a known fact that voice assistants can irritate when calling a bank or an online shop, however, it shows that with a positive experience of communicating with a voice assistant, customers start to use it regularly to acquire necessary information, because waiting infinitely to receive the information can irritate them even more. Moreover, voice assistants exclude the possibility of a human error while consulting a customer (and these often happen). Positive experience can also be created with less “dull” voice technologies, so, for our customers, we develop maximally “human” voice assistants to communicate, based on fine tuning of voice parameters.
- Audiobooks. But the usage if voice technologies is not limited by those directions and there are a lot of spheres where Voice AI can transform everything. For example, now we are developing a cool product to dub audiobooks in English and Russian. In this case technology helps a human to dub an already prepared text. In its turn, a human can correct stresses, change the mood of narratives and use different voices. As in the examples above, AI solves here the most labour-consuming task. Taking into account that the audiobooks market is constantly growing (20%+ CAGR), and bloggers find it trendy to dub their articles (Medium implemented this feature several years ago for the most popular articles), this technology will become highly sought in the next decade.
- Robots and IoT. Active attention has lately been attracted to robotics and the internet of things, where voice assistants are a key feature. Until now, one of the problems while developing them was their inability to work without the Internet — we have solved this problem developing a self-hosted version of an engine, and thus, voice assistants can be used in manufacturing and other places where there might be problems with the Internet. It is these innovations why IoT devices have in the last several months flooded all industry in the world (especially, mining).
Our experience shows that the right usage of Voice AI opens new opportunities for various markets — for example, our current challenge is to master a big number of languages (last year there was a success story of selling an Indian start-up which can recognize ten local languages). That is why we think that, despite the hype around AI in the past decade, its real value is beginning to uncover only in the present one.
