By Yashar Ahmadpour and Dimitry Fisher
Artificial Intelligence (AI) has been all the rage in the last few years in tech as well as business, and it will continue to at an increasingly rapid pace. Turn the pages of any tech magazine, blog, or even a newspaper, and you are bound to read a story about AI in one form or another. But is A.I. something new, or is it older than what most people know or think?
The term “Artificial Intelligence”, for computers trying to solve problems as a human would, was adopted in 1956 “when the first academic conference on the subject was held.” However, centuries earlier, machines (“automata”) were already designed that could perform some human tasks: “…Da Vinci may have actually built a prototype in 1495 while working under the patronage of the Duke of Milan.” Should we look to the cinematic arts, then there is no greater example than 1927 movie Metropolis by German filmmaker Fritz Lang. Steven Spielberg’s AI Artificial Intelligence was released in 2001, which was based on a 1969 story. But, what all of these movies get wrong, is that most of them depict AI as a robot that is replacing humans when AI is really here to enhance the human workflow. It is here to help our work get done with greater accuracy, and more efficiently. It allows for more meaningful work to be done by humans while AI performs menial repetitive work.
In the Industrial Revolution, when we transitioned to new manufacturing processes, entire new industries were built, propelling humankind forward. And those manufacturing processes just kept improving over time. We had, and still have the same going on right now with the communication networks and the Internet. And the same can be said about AI, as the work of AI is never done. To understand why that’s the case, all we need to do is to look to how we raise children. An infant may start to learn new words at about 12 months, and those words are associated with objects. A prime example of this is when you are attempting to teach a child what a balloon is. You show an actual balloon to a child and say the words balloon while pointing to the balloon. After many repetitions of the word and the association of the object, the child will learn to associate the word balloon with an actual balloon. When an AI learns, too, often times this is exactly what is done. (not all AI agents learn, but most modern ones do. This approach to AI is known as Machine Learning). When a toddler is shown a soccer ball, they may call it “balloon” as well. The machine, too, may identify the soccer ball as a balloon, or simply fail to identify it at all. And this is where you as a parent teach your child that what they are in fact looking at, is a soccer ball, and not a balloon. Data Scientists will do the same, except rather than point to a soccer ball and utter the words soccer ball, they would label and show thousands of images of soccer balls in different lighting, environments, and more. That is an example of what is known as supervised learning. An improved, semi-supervised learning approach may mix labeled examples with many unlabeled but similar ones. As children’s drawings provide a glimpse into how they see the world, an AI too can “imagine” or generate images using elements or features it has learned (as it is, mostly puppies). Like the children improve their skills by playing, two AI agents may improve each-other by competing: one draws increasingly more accurate pictures of, say, soccer balls, and the other gets increasingly better at telling those drawings from the images of the real thing. As an AI agent keeps learning new things, new things, new tasks, and new approaches are created every single day. This is why the job of AI is never finished.
The AI agents of today are primarily intended to solve specific well-defined data-based tasks. Children have real people and real world to interact with and learn from year after year, not to mention half a billion years of brain evolution. By contrast, an AI’s entire world is the dataset it’s trained on. (Decision-makers may want to read more about the importance of this here). It is the future generations of AI that will be more like our children, kinder and smarter than we are, living alongside us and learning from us.
If you are in the market for learning more about AI or are simply curious about it, feel free to reach out to us. We are always happy to talk about AI and Machine Learning, and if you are in need of AI but don’t know where to start, then we are here to guide you. You could also prep yourself by reading our past post, Your AI Checklist: 5 critical steps to determining if your business is AI ready to get ready for AI.
About the Authors
Yashar is a 3x startup founder, who raised capital for all three companies, where he built, and shipped the products. His last startup was based in the Bay Area, where he was joined by famed VC Neil Weintraut of Palo Alto Ventures as his co-founder. Since 2017, at the birth of his daughter, he decided to focus his time and efforts in San Diego where he lives with his family. Since then, Yashar has joined Analytics Ventures where he leads Product and works with a brilliant team of AI Scientists, and Engineers.
In his spare time, Yashar loves to spend time with his family, going to the Zoo, the beach, and traveling as often as possible. He has also joined forces with other San Diegans to form San Diego Tech Hub to provide greater awareness and draw to the San Diego Tech Scene.
Dimitry Fisher is a Chief AI Officer at Analytics Ventures / Dynam AI. Dimitry has 20+ years of R&D experience in academia, government, and private sectors, spanning multiple areas of physics, neuroscience, ML and AI. Dimitry received his Ph.D. in Physics from Weizmann Institute in 2002. He did independent research in physics before shifting his focus to neuroscience and to mechanisms of learning in neural networks. Dimitry joined Brain Corporation in 2012, doing bleeding-edge R&D for Qualcomm, DARPA, and several internal projects, including the first self-driving janitorial AI. Dimitry saw Brain Corporation grow from 10 people to 100 before resigning his position to join Analytics Ventures as the Chief AI Officer. He has 12 granted US patents and over 40 publications in scientific journals.