Sunday, December 13, 2015

OpenAI: AI for All


Yesterday saw the announcement about the birth of OpenAI, a non-profit organization to develop and advance Artificial Intelligence (AI) technologies, and share these in the greater good.  Infosys, and I, are a part of this endeavor, and very excited about it and I've been asked tons of questions in the last 24 hours about this, so I thought I'll write some thoughts down.

Why?

A few weeks ago, Marvin Minsky, one of the fathers of Artificial Intelligence (AI), the one who gave the field its first definition -- that AI is "the science of making machines do things that would require intelligence if done by men" -- made a few sobering statements about the state of the art in AI.  Indeed I felt sad listening to this giant lament the lack of fundamental progress in the field, and highlight some of the underlying causes.  And this despite all the buzz and hype AI work has picked up lately.  Marvin is one of the truly great human beings and scientists, whose teachings and advice helped influence my life and led me to focus on AI in my grad studies and beyond.  So more than anything else, I see OpenAI as an opportunity to "do something about it".

My friend and teacher Alan Kay once referred to Sam Altman as a "builder of civilizations".  When Sam, a wise man who is but 30 years old, was thinking about the idea of building an open ecosystem for, among other endeavors, AI, Alan and I shared our ideas with him and our experiences.  Sam asked me if I would be ok with the fact that such an endeavor would be untethered and would produce results generally in the greater interests of humanity, and he was somewhat surprised by my reaction, that indeed I would only support this venture if such an openness was a fundamental requirement!  In all my experience with corporate research teams, I found a continual struggle for the teams to find relevance with the work in the "here and now", usually knowing that this unnecessary and premature seeking of relevance not only blinds us to those opportunities that can shift our paradigms, it defeats the point of research.  He shares the view that cooperation helps dramatically improve our lot, helps create a foundation for much larger value creation than any isolated "feudal" system can.  Indeed, endeavors such as agriculture, and science, show that when we share, we improve all of us.  As Newton once said, "if I have seen further, it is by standing on the shoulders of giants".

I am really excited that Sam, Elon Musk, and others -- including Reid Hoffman, Peter Thiel, and Amazon Web Services, and Infosys of course -- are supporting this great endeavor, and in addition to Alan, a great set of leaders will serve as advisors.  Ilya Sutskever, who has worked over the last several years as a leader in developing the so called "deep learning" techniques, will direct the research at OpenAI, and will be joined by several dynamic leaders and individual contributors, whose conviction and imagination will help this field move forward, based on the best of what we know, the best ideas, the best inventions, and the key lessons.

Our wish is that together the OpenAI team will do unfettered research in the most important, most relevant dimensions of AI, no matter how long it takes to get there, not limited to just identifying dancing cats in videos, but to creating ideas and inventions that amplify our humanity, that help us learn more, see/perceive and understand more, and be more.

Why Open?

One question that's been asked since yesterday, is why should this be open?  Isn't it better to have deep AI be in the hands of a select few experts or specialists?  My sense is, our trust in complex systems stems mostly from understanding these and their predictability, whether it is nuclear reactors, lathe machines, or 18-wheelers; or of course, AI.  If complex systems are not open, not open to be used, extended, and learned about, they end up becoming yet another mysterious thing for us, ones that we end up praying to and mythifying.  The more open we make AI, the better.

Why Infosys?

Another question that's been asked a lot, is why Infosys?  We at Infosys, with over 150k software engineers, are unique beneficiaries of and contributors to this endeavor.  Most of our work is in building and maintaining software systems, and AI will increasingly shape the construction and evolution of intelligent software systems, in all kinds of domains and industries.  In addition, as a large services company, many parts of our work can transform fundamentally with AI.  In services like infrastructure management, business process outsourcing, and verification and maintenance of existing software, we can massively migrate mechanizable work to automation, and instead build intelligent software systems, that amplify us, our abilities, as well as those of our customers.  So a great transformation that we are undertaking at Infosys, is to embrace automation at a very large scale, so people can, as Prof Mashelkar once said, "do more with less for more", and at the same time, educate ourselves in new areas to help build intelligent systems, but also to innovate in our work, to exercise our creativity in everything we do, and amplify our abilities, our humanity, using AI.

But beyond business, there is another key reason; our endeavor to do purposeful work.  Our founders always believed in this.  Many years ago, Mr Murthy and our founders started the ACM Infosys award, which celebrates great young Computer Science practitioners.  The Infosys Science Foundation supports work in the pure sciences, as does the Infosys Foundation in India.  And most recently, the Infosys Foundation in the US, is working hard on its mission to help enable/expand computer science education, and has already started many promising initiatives here in the US.  So OpenAI aligns very nicely with our long-held values.

So as we get started on this great journey, I find myself excited, hopeful that OpenAI will help uncover great innovations, that new AI techniques yet to be discovered, and built to share, built in ways that are open to all, will help us transcend our limitations, improve and amplify us all, and that our work in artificial intelligences, may help bring us closer to our natural spirituality...

-Vishal
Palo Alto,
Dec 12, 2015