The Race Towards Democratizing Artificial Intelligence
Artificial Intelligence (AI) was born in the summer of 1956 at the Dartmouth Conference, where various scientists from mathematics, psychology, engineering, economics, and political science discussed the possibility of having an artificial brain. The conference introduced the Logic Theorist program – the first of its kind that copies humans’ problem solving skills. It was then that John McCarthy convinced his peers to accept AI as the name of the field.
Since then, AI has strived to build machines that can:
- interact with the real world;
- plan and navigate the world that humans live in;
- speak natural human language;
- be “taught” how to perceive things; and
- create a fully autonomous computer that can continuously learn by itself.
The Most Notable AI Acquisitions
Fast forward to 2016 – Google, IBM, Facebook, and Baidu are leading the game towards AI. A study was conducted by CBS Insight this year showing all major acquisitions done by these tech giants catering to various industries.
Diagram 1: Major Acquirers in Artificial Intelligence
Source: CB Insights
Acquisitions have gained momentum since 2011 and continued to peak – much like a race towards AI leadership. Since then, 140 private companies were bought, in which Google remains the top purchaser. Its reorganization in 2015 led to a new entity called Alphabet, leading the tech giant to systematically diversify its verticals hence founding Calico for healthcare; Google X for robotics and self-driving cars; and Nest for smart homes. Perhaps Google’s acquisition of Boston Dynamics in 2013 and DeepMind Technologies in 2014 paved way for the serious groundwork of its AI initiatives.
Race Towards AI Dominance
It was DeepMind’s work that beat a human world champion in the board game “Go.” Considered as an AI milestone, Google beat world champion Lee Sedol on this famous East Asian game that is so complex due to the boundless possibility of moves, each requiring serious strategic thinking. Unlike its AI predecessors – Deep Blue (1997), IBM Watson’s Jeopardy – AlphaGo won due to its machine learning capabilities. Deep Mind combined Monte-Carlo tree search with deep neural networks formed from reinforced and supervised learning from human game experts.
Other major acquisitions also happened by Intel and Apple. Intel bought Itseez, Nervana System, and Movidius; Apple absorbed Turi and Tuplejump. Intel’s acquisitions are geared towards robotics and machine learning aspect of AI. Itseez, for example, focuses on machine vision, or the capacity for computers to view and understand their surroundings. This tech is mainly used for autonomous cars, robotics, and even security. Apple on the other hand harnesses machine learning, with its acquisition of Tuple Jump which provides data mining and analysis. Turi is also a machine learning platform for developers and data scientists allowing sentient analysis on social media. Last year, Apple also acquired VocalIQ, which aims to make it easier for robots to speak.
AI for healthcare is in demand. Acting as an adviser, AIs can augment healthcare professional advice through tapping into a massive amount of patient data. IBM remains on top of the AI healthcare game. Watson uses natural language processing, information retrieval, knowledge representation, automated reasoning, and machine learning technologies to provide question answering (QA) services. Watson made its debut on the quiz show “Jeopardy!” in 2011 where it won first place. In the same year, Watson was used commercially at Memorial Sloan Kettering Cancer Center to help in lung cancer treatment decisions in partnership with WellPoint. Now, Watson is the most widely used tech in healthcare and has partnered with Mayo Clinic, CVS Health, and New York’s Memorial Sloan Kettering Cancer Center (MSK) just to name a few.
Facebook also has its own initiative called Facebook AI research (FAIR) focusing on machine intelligence that will enable better communication. One of its research outputs, called fastText, is an open-source text classification tool based on the skip-gram model. As its authors state, “In order to be efficient on datasets with a very large number of categories, fastText uses a hierarchical classifier, in which the different categories are organized in a tree, instead of a flat structure.” Facebook expects that fastText will help in curtailing spam and quick bait filtering because the tech can be quickly trained to use extremely large datasets.
How Far Are We from Skynet?
“Google is Skynet” is on top of the suggested results from their own search engine. Every time speculative thinking on AI is brought up, it culminates in a post-apocalyptic reality where a fictional neural net-based AI and conscious group mind system that is Skynet dominates humanity. In the movie Terminator, Skynet is supposed to help the U.S. military reduce mistakes and make quick decisions. However, upon gaining artificial consciousness, it arrived at a logical conclusion that humans must be destroyed to fulfill its mandate of safeguarding the world.
Humor aside, we can actually say that Skynet is not pure fiction. The technological singularity hypothesis states that eventually artificial intelligence will pave way to a “runaway reaction” of self-improvement cycles – meaning that machines will start improving by themselves, fully independent of man. This plausibility is suggested by Moore’s Law , stating that technological change increases exponentially, including integrated circuits, improved to transistors, vacuum tubes, relays and eventually electromechanical computers. These iterations will continue exponentially and in a few decades, computers can exceed human brains.
CB Insights released a study this year showing an AI heatmap grouped by industry stating where artificial intelligence is heavily utilized. Leading the game is healthcare, followed by advertising, sales and marketing, and business intelligence.
Diagram 2: Artificial Intelligence: Sub-Industry Heatmap
Source: CB Insights
Using this heatmap, we’ve identified the initiatives of two tech leaders who have a robust AI strategy – Google and IBM. This matrix can help developers and engineers identify which technology to use when starting to build using AI technology stack.
The Future of AI
Tech giants are investing, continuously acquiring, and supporting almost all AI initiatives geared towards the goal of democratizing it, fundamentally changing how humans interact with machines. As developers, we can watch out for emerging tech brought by all of these acquisitions and seek how to utilize them in our existing projects. One thing is for sure – the work on AI will continuously improve and is speculated to reach human-like reasoning before 2100. This claim is not your typical “Google is Skynet” meme; it is in fact a feasible argument stemming from Moore’s Law. The heatmap above shows the industries where AI initiatives can be fruitful and our matrix can help you identify which technology stack to use.
Image compliments of meme generator
For more articles that will fuel your fire of web-possibilities, click the button below and subscribe to this blog.