The Google AI ecosystem consists of a range of applications and platforms that have artificial intelligence (AI) and machine learning (ML) incorporated into their underlying technologies and infrastructure. But does this make these products AI, in and of themselves?
That’s what we’ll be exploring in this article, by looking at the functionality and behind the scenes workings of some of Google’s most popular offerings.
First, though, we need to establish exactly what it is that we mean by AI.
What Is Artificial Intelligence (AI)?
AI or artificial intelligence is a blanket term for a branch of computer science bringing together numerous disciplines, with the objective of producing machines and systems that can do things previously only possible through the application of human reasoning, learning, or perception.
Strong AI or artificial general intelligence (AGI) is the long term goal of many in the field and represents a state where intelligent machines will be able to perform tasks in a manner equal to or surpassing the capabilities of human cognition. Ask most analysts, and they’ll tell you that this condition is years, if not decades, away.
For the moment, the artificial intelligence we have is narrow or weak AI — systems designed to perform a specific task or limited range of tasks very well.
What Is Machine Learning (ML)?
Machine learning or ML is itself a sub-branch of artificial intelligence. It involves collecting and processing huge amounts of data, identifying patterns and links within this data that wouldn’t immediately be apparent to humans, and the use of those observations in making decisions and improving the operations of the system. Machines of this type can effectively “learn” from the data they collect and from past or present activities and interactions.
What Constitutes Artificial Intelligence?
Opinions on what actually qualifies as artificial intelligence will differ, depending on who you talk to. In the early stages of AI development, the best the field had to offer were expert systems — an attempt to emulate human knowledge by programming extensive rules into computers. With the development of machine learning technologies, this early class of machines is no longer considered by many to be true artificial intelligence.
According to some classical definitions, AI requires “a flexible rational agent that perceives its environment and takes actions that maximize its chance of success at an arbitrary goal.” Artificially intelligent systems must be constructed as rational agents, capable of understanding what a question is — and understanding its answer. In strict terms, such technology also has the ability to refuse to answer a question, based on some internal criteria for which it has determined that answering the question will negatively affect one of its goals.
Is Google an AI? Specifically, is Google Search Artificial Intelligence?
Historically, the Google Search engine was driven by algorithms (complex mathematical procedures spelling out the parameters governing a system’s function) that automatically generate a response to each query. Essentially, these algorithms amounted to a set of definite rules, which Google engineers could readily change and refine.
Over time, Google Search has been incorporating machine learning and deep learning technologies — an advancement on ML that uses artificial neural networks modeled on the human brain and nervous system. It’s been an additive process, and overall, Google’s Search Algorithm is now a combination of various algorithms that serve different purposes.
For example, the Hummingbird Algorithm introduced Semantic search and Natural Language Processing (NLP, a system enabling better communication between humans and machines), which allowed Google Search to understand the intent of a search query, rather than having to rely on exact keywords.
One of the major advancements in Google Search over the past few years is Voice Search, which helps to fine-tune the search database with newer keywords, semantics, sentiment, and other factors.
But if you ask Google Search a question in its current state of development, it won’t refuse to answer your query because of some internal conflict. Google Search is more of an expert system that’s becoming increasingly versatile through the use of machine learning.
Is Google Assistant Artificial Intelligence?
Launched in 2016 as an extension of Google Now, Google Assistant was designed to be personal, while expanding on Google’s existing “OK Google” voice controls. Assistant fuses the personalized elements offered by the previous system (such as knowing where you work, your favorite sports franchises, and your meeting or travel plans) with a wide range of input controls. The platform supports both text or voice entry and will follow the conversation whichever entry method you use.
Once you activate the system with the “OK Google” or “Hey, Google” wake words, Google Assistant offers voice commands, voice searching, and voice-activated device control, enabling you to complete a number of tasks including:
- Controlling your devices and smart home
- Getting instant access to personal and calendar information
- Finding information online
- Organizing and playing your music
- Running timers and reminders
- Making appointments and sending messages
- Opening apps on your phone
- Reading your notifications to you
- Playing games
- Making real-time language translations
Once you’ve initiated the system, Google Assistant listens for a response without needing a trigger phrase all the time. It can also recognize voice profiles for different people and tailor its response accordingly.
At its heart, Google Assistant is an AI-powered system that ties together a range of tools and services from first- and third-party brands to create new methods for running apps and controlling the smart home, enabling Google to automate and anticipate the wants and needs of consumers. But with its current limited scope to act autonomously, it’s not a pure AI.
Is Google Translate Artificial Intelligence?
Advances in machine learning (ML) have been driving improvements to automated translation, including the GNMT neural translation model introduced in Google Translate in 2016. This enabled the platform to improve the quality of translation for over 100 languages greatly.
For high-resource languages like Spanish and German, which have plenty of training data available, the system has traditionally worked very well. But performance for low-resource languages (which are often spoken by limited or widely dispersed populations) translation quality has been low and unavailable in some cases.
In 2020, Google is launching five additional languages for Google Translate, bringing its total number of supported languages to 108 — its first significant addition to the roster in four years. Recent advances in machine learning and help from members of the Google Translate Community have enabled it to provide coverage for Kinyarwanda, Odia, Tatar, Turkmen, and Uyghur, around 75 million people collectively speak languages. Google says that the four-year delay was due to the interval required by its machine learning systems to assimilate the new data.
Besides the consumer level app, which allows users to translate text in other languages for free, Google offers the Translate API and AutoML Translation — products that cater to businesses with little machine-learning expertise, enabling them to build custom translation models.
But these state-of-the-art systems still lag significantly behind human performance in all but the most specific translation tasks. Digital models are still prone to typical machine translation errors, such as poor performance on particular types of subject matter, different dialects of a language, producing overly literal translations, and poor performance on informal or spoken language.
Google Translate is an expert system that’s still far from perfect.
Is Google Maps Artificial Intelligence?
Of all the products in the “Google AI” stable, Google Maps has possibly the strongest claim to artificial intelligence. And with features like Driving Mode, which estimates where you are headed and helps you navigate without any commands, Google Maps is also one of the company’s most widely-used products. Thanks to machine learning tools from DeepMind, the London-based AI lab owned by Google’s parent company Alphabet, its features have become even more accurate.
Google and DeepMind researchers take data from a variety of sources and feed it into machine learning models to predict traffic flows. This information includes historical traffic data, live traffic information collected anonymously from Android devices, data provided by local governments (speed limits, location of construction sites, etc.), and each road’s physical characteristics.
Neural networks designed by DeepMind extract patterns from this data and use them to predict future traffic. Data modeling works by dividing maps into “supersegments” — clusters of adjacent streets that share traffic volume. Each is paired with an individual neural network that makes traffic predictions for that sector. Each supersegment changes in size dynamically to match traffic flows, and (according to Google) draws on “terabytes” of data.
It’s an adaptive system fueled by information which also enables Google Maps to learn from events and usage.
Is Google Home AI?
The Google Home approach to smart home management is based on Google’s “Helpful Home” strategy, which was unveiled in 2019. Rather than being a single AI entity, Google Home is more of a collection of AI-powered systems like Google Assistant, which serves as a coordinating medium for the system.
Google Home the brand, is actually a Chromecast-enabled speaker, which serves as a voice-controlled assistant providing Google Assistant functionality in the home. It’s part of a range that includes the Google Home, Google Home Max, Nest Mini, the Nest Hub, and Nest Hub Max.
Using voice control, you can ask Google Home devices to do anything that you’d ask Assistant to do on Android phones, with an added emphasis on in-house systems such as lighting, heating, security, and entertainment.
Google Assistant (and therefore, the Google Home ecosystem) currently works with over 1000 home automation brands and more than 10,000 devices. And the range of Google Home applications is increasing due to features like Google Assistant Connect — a platform that device manufacturers can use to bring the Google Assistant into devices more easily and at an affordable rate.