The Consumerization of Artificial Intelligence
Consumerization is the design, marketing, and selling of products and services targeting the individual end consumer.
Apple CEO Tim Cook recently promoted a $100-per-year iPhone app called Derm Expert. Derm Expert allows doctors to diagnose skin problems using only their iPhone. Doctors take a photo of a patient’s skin condition and then Derm Expert diagnoses the problem and prescribes treatment. Doctors can effectively treat patients without a high performance computer or an expensive technology environment. They just need the same iPhone that you and I use every day.
Derm Expert makes use of Apple’s Core ML framework that is built into all new iPhones. Core ML makes it possible to run Machine Learning and Deep Learning algorithms on an iPhone without having to upload the photos to the “cloud” for processing.
Apple is not the only company integrating Machine Learning and Deep Learning frameworks into their products, but it may be the first company to put such a powerful capability into the hands of millions of consumers. Whether we know it or not, we have all become “Citizens of Data Science,” and the world will never be the same.
Embedding Machine Learning Frameworks
Apple Core ML in the iPhone is an example of how industry leaders are seamlessly embedding powerful machine learning, deep learning, and artificial intelligence frameworks into their development and operating platforms. Doing so enables Apple IOS developers to create a more engaging, easy-to-use customer experience, leveraging Natural Language Processing (NLP) for voice-to-text translation (Siri) and Facial recognition. Plus, it opens the door for countless new apps and use cases that can exploit the power of these embedded frameworks.
Core ML enables developers to integrate a broad variety of machine learning algorithms into their apps with just a few lines of code. Core ML supports over 30 deep learning (neural network) algorithms, as well as Support Vector Machine (SVM) and Generalized Linear Models (GLM).
- Developers can integrate computer vision machine learning features into their app including face tracking, face detection, landmarks, text detection, rectangle detection, barcode detection, object tracking and image registration.
- The natural language processing APIs in Core ML use machine learning to decipher text using language identification, tokenization, lemmatization and named entity recognition.
Core ML supports Vision for image analysis, Foundation for natural language processing, and GameplayKit for evaluating learned decision trees (see Figure 2).
Machine Learning and Deep Learning Microprocessor Specialization
Artificial intelligence, machine learning and deep learning (AI | ML | DL) require massive amounts of computer processing power. And while the current solution is just to throw more processors at the problem, eventually that solution won’t scale as the processing needs and the volume of detailed, real-time data increase3.
One of the developments leading to the consumerization of artificial intelligence is the ability to exploit microprocessor or hardware specialization. The traditional Central Processing Unit (CPU) is being replaced by special-purpose microprocessors built to execute complex machine learning and deep learning algorithms.
- Graphics Processing Unit (GPU): a specialized electronic circuit designed to render 2D and 3D graphics together with a CPU. It is also known as a graphics card in the gamer’s culture. Now GPUs are being harnessed more broadly to accelerate computational workloads in areas such as financial modeling, cutting-edge scientific research, deep learning, analytics, and oil and gas exploration etc.
- Tensor Processing Unit (TPU): a custom-built integrated circuit developed specifically for machine learning and tailored for TensorFlow (Google’s open-source machine learning framework).TPU is designed to handle common machine learning and neural networking calculations for training and inference, specifically: matrix multiply, dot product, and quantization transforms. On production, AI workloads that utilize neural network inference, the TPU is 15 times to 30 times faster than contemporary GPUs and CPUs, according to Google.
Intel is designing a new chip specifically for Deep Learning called the Intel® Nervana™ Neural Network Processor (NNP). The Intel Nervana NNP supports deep learning primitives such as matrix multiplication and convolutions. Intel Nervana NNP enables better memory management for Deep Learning algorithms to achieve high levels of utilization of the massive amount of compute on each die.
The bottom-line translates to achieving faster training time for Deep Learning models.
Finally, a new company called “Groq” is building a special purpose chip that will run 400 trillion operations per second, more than twice as fast as Google’s TPU.
What do all these advancements in GPU and TPU mean to you the consumer?
“Smart” apps that leverage these powerful processors and the embedded AI | ML | DL frameworks to learn more about you to provide a hyper-personalized, prescriptive user experience.
It’ll be like a really smart, highly attentive personal assistant on steroids!
The Power of AI in Your Hands
Unknowingly over the past few years, artificial intelligence worked its way into our everyday lives. Give a command to Siri or Alexa and AI kicks in to translate what you said and look up answer. Upload a photo to Facebook and AI identifies the people in the photo. Enter a destination into Waze or Google Maps and AI provides updated recommendations on the best route. Push a button and AI parallel parks your car all by itself (dang, where was that during my driver’s test!).
With advances in computer processors and embedded AI | ML | DL frameworks, we are just beginning to see the use cases. And like the Derm Expert app highlights, the way that we live will never be the same.
Figure 2: Core ML