AI asks – no, forces – us to embrace new design patterns.
Founded in 2009 at Orange Silicon Valley as a way to engage with Stanford and Cal, Orange Institute has iterated over 18 sessions into a global conversation platform connecting brands with science and entrepreneurship. Orange Institute is an ongoing expression of what founder and honorary president Georges Nahon calls “pragmatic altruism” — the belief that we learn better together with others.
For three very intense days in April, 45 Institute members from 35 companies crisscrossed Silicon Valley to engage with an interdisciplinary faculty consisting of data scientists, designers, entrepreneurs, and academics with a common focus: machine intelligence. Venues included the Jacobs Institute for Design Innovation on the UC Berkeley campus, the Visa Innovation Center at One Market in San Francisco, the RobotX accelerator in Santa Clara, and the Autodesk Gallery across the street from Orange Silicon Valley’s offices.
Holding on to the fils rouge of machine intelligence as our thread, the multi-disciplinary group of CXO’s, digital strategists, and innovation execs met the smartest people in the Valley working either as enablers – think Nvidia and Amazon Web Services – and in verticals ranging from adtech to agtech to connected cars and cubesats.
“The opportunity for IO #18 was to show how horizontal machine learning tech, like computer vision, is popping up in multiple verticals,” noted Mark Plakias, the event’s lead curator from Orange Silicon Valley.
Machine learning is a very big tent. Indeed, Nvidia exec Kimberly Powell opened the session by asserting AI is the defining model for IT going forward. Amazon Web Services PM for AI Joe Spisak drove the point home by demonstrating the resources AWS is putting forth for developers in the cloud, and in the open source community with its aggressive sponsorship of Apache MXNet to counter anxiety about the enclosure of AI. Uber’s former AI/ML czar Danny Lange brought the infinite patience of algorithmic learning to life with vivid animations of AIs learning everything from Pong to Go to virtual chickens crossing the road.
AI’s progress towards mastery took center stage at the sparkling new design hub at Cal funded by Paul Jacobs of Qualcomm. Anca Dragan, a Romanian roboticist and charter member of the Center for Human-centered AI at UC Berkeley, led by AI guru Stuart Russell, brought into light the nexus of objectives we give these algos by reminding us of the King Midas parable.
Dragan’s discussion of how we reward reinforcement learning machines complements the other AI policy analysis from Open AI’s strategy director Jack Clark, who reports to the founders Elon Musk and Y-Combinator CEO Sam Altman. Clark reminded us both of the limitations of these algorithms, as well as their potential to go horribly awry, based on improper goal-setting. In both cases, the ability of humans to override their algorithmic creations is now an operative research issue — and policy embodied in code.
More immediately pressing on the planning horizons of many CMO members was chatbots – and what to do beyond the hype. Orange Silicon Valley expert David Martin conducted a master class with bot platform evangelist and Pandorabots CEO Lauren Kunze, and medical avatar creator and entrepreneur Adam Odessky of Sense.ly.
By the end of day one it was clear that machine intelligence is not just a new IT paradigm but – more fundamentally – an epochal design challenge across humanistic as well as technical planes. From the IT perspective of how GPUs drive edge computing and machine learning at the edge, to the policy implications of personal data collection as explored by Dan Elitzer from IDEO, we are being asked – no, forced – to embrace new design patterns.
The focus over the balance of the session moved towards commercial and industrial aspects of these design challenges. As the group moved into high gear, we would run the gamut from farm tractors to self-driving cars to putting fleets of cubesats into space.
Jonathan Salomon from Orange Silicon Valley gave the roadmap to autonomy for vehicles, showing how cars can reach full Level 5 autonomy over the next decade. The changing landscape of automotive is exemplified next by NIO – a Chinese-funded EV play with a US HQ stocked with people like Chris Pouliot, a bona fide data scientist. Pouliot sees every aspect of the car as a software-controlled platform, optimized and personalized.
If NIO is building it, Karen Kaushansky is designing it. She is already past working on the question of intent for robots, she is working on how robots communicate their intent to humans.
Just how cities adapt to autonomous vehicles, and how an increasingly urban population thrives in the age of digital intelligence takes us into once-disparate but now AI-connected worlds. Who knew that the future of farming implicates satellites? Thanks to swarms of camera-equipped small cubesats Mother Earth can have her picture taken every 24 hours, yielding crop predictions at a macro scale. Here on earth and close-up, the same computer vision tech can drive through vineyards to count individual grapes for yield predictions.
The takeaway from these canonical physical domains of city planning, farmlands, and space satellites, is that the value has shifted from the standalone physical, to the connected and digital. Where entrepreneurs win now is where the data emanates from the physical.
“Value moves away from owning assets and toward what you do with data.” –Eric Anderson, AndOne Technologies
The best part of bringing sophisticated, accomplished executives to Silicon Valley is watching them apprehend the intensity of startups and corporate innovation teams. Here at Orange in SF, we are blessed to have two global centers of innovation on our block. So we close the three-day immersion by literally crossing the street to spend time at the Autodesk Gallery, and the Visa Innovation Center. Close to our home at the Embarcadero, we enjoyed inspirational talks by our corporate neighbors, as well as startups that are emblematic of the new model: AI in everything.