A 6 minute Intro to AI
“I know what AI means... I think”
Between Hollywood and your dusty stack of sci-fi novels, you’ve been given many outlandish representations of the AI of the future. But AI is already here. It’s all around us — just in humbler forms.
Intelligent behavior in an autonomous agent — THIS is AI. It’s describing the brain, not the body, of intelligent machines (AI ≠ robots). The AI of today can do specific tasks (driving a car, booking meetings, picking your next Netflix binge). AI research is leading toward something more advanced: artificial general intelligence, or AGI. This AI — when a machine can do things in a way that is indistinguishable from human behavior — is what we’re all waiting for.
The Symphony of AI
AGI « Artificial general intelligence », is the Grand Finale at the end of a symphony. But before we strike that last glorious chord, each individual instrument must be played with great expertise. At Snips we believe the cornerstone instruments of AI include: Machine Learning, Deep Learning, Natural Language Understanding, Context Awareness, and Data Privacy. Play on!
Machine learning and AI are not the same. Machine learning is an instrument in the AI symphony — a component of AI. So what is Machine Learning — or ML — exactly? It’s the ability for an algorithm to learn from prior data in order to produce a behavior. ML is teaching machines to make decisions in situations they have never seen.
The most mainstream approach to ML is showing the algorithm a data set of situations and telling it what the right decision is — training a model. This is supervised Machine Learning. Once the model has been trained, we can feed new, more foreign data through the algorithm — and hopefully, the machine makes intelligent decisions in these new, foreign situations.
Deep learning is a branch of machine learning where artificial neural networks — algorithms inspired by the way neurons work in the brain — find patterns in raw data by combining multiple layers of artificial neurons. As the layers increase, so does the neural network’s ability to learn increasingly abstract concepts.
For example, neural networks can learn how to recognize human faces. How? The first layer of neurons takes pixels from example images, the next layers learn the concept of how pixels form an edge, then that layer passes that knowledge to other layers, combining that knowledge of edges to learn the concept of a face. This process of layering knowledge continues until BAM! — the neural network algorithms recognize specific features, and ultimately specific faces.
AI must communicate with humans as well as humans communicate with each other. In AI, this level of understanding is called Natural Language Understanding, or NLU. NLU is a huge priority and challenge in AI research. Why? Because human communication is not straightforward. It’s a complex web — random, out-of-order, peppered with humor, emotion and conflict — and it depends hugely on context.
Once AI conquers the challenge of human communication, decoding complex questions (natural language queries), making connections, and giving answers that make sense, radical progress is not far behind.
Like a human assistant, an AI assistant can only be as smart as the information — the context — you give it access to. If your assistant — human or artificial — only has the ability to see your calendar and reservations, but not your contact list and location data … well, that’s not a very helpful assistant.
Context is king when it comes to complex tasks. It’s true of humans and it’s true of AI. Every section of data and context needs to be tuned perfectly to play a different note in the symphony of AI.
Rise of Voice Commands in Assistants
As the quest for natural language communication continues, speech-to-text technology has improved immensely. A new-and-improved Siri and the launch of Amazon Echo and Google Home are prime examples of this science fiction storyline coming true.
AI isn’t the Rise of the Machines, it’s the Machinification of Humans
While AI movies and TV feature robots with human bodies, many fail to explore AI (tiny, tiny, TINY AI robots) IN human bodies. AI visionaries like Elon Musk are starting to talk AI-human symbiotes, with AI nanotechnology effectively curing humans of … death! Sounds like a good storyline.
Content as a Testing Ground
In order to get smarter, AI requires lots of data, patterns and new situations. Enter content platforms. Users’ consumption patterns are already being shaped by the machine learning behind Spotify’s “Discover Weekly”, Netflix’s “Recommended For You” and Facebook’s ability to keep you in a filter bubble of your own making.
AI is learning to be less biased
Technology is both inescapable
We reached a zenith of tech inescapability in the last few years. However, this same period also marked a widespread response to over-tech-ing … the birth of several “mindful technology” movements and products — witness Tristan Harris’ Time Well Spent and the 24-hour holiday National Day of Unplugging, to products like The Light Phone.
For any of this to happen,
we need Data Privacy
The future of AI is dependent on data privacy. Why? Because without data to learn from, AI cannot get smarter and progress will slow. Companies must commit to creating private and secure products. Users need to know that their personal data will be protected if they’re ever going to permit an AI full access.
Why are we sharing this?
Snips has put years of hard work and research together to build AI powered Voice technology that allows anyone to build a powerful voice assistant into their product or device. We want to create AI that is so good, it can eventually make technology disappear. But that’s not all.
Living in the world of AI like we do in Snips, we hear the same questions over and over. You deserve answers. So here it is — a quick tool to empower anyone to better understand the basics of AI, because after all, it’s all around us. Sharing this knowledge matters.