Google[x]: Building a Moonshot Factory

Moonshots are seemingly impossible and yet impossibly-important ideas that through science and technology can be brought to reality. Google[x] is a moonshot factory full of optimists who are focused on changing the world by seeking out massive unsolved problems that — when solved — will profoundly and positively alter the way we live. You may have heard of self-driving cars and Google Glass, but here we’ll give a glimpse of the ethos, style, and people behind Google[x].

Then, from 4-6pm, come to our “Solve for X” exploration session at Bat Bar to engage in pushing forward moonshots — radical technology-based proposals for solving global problems (register at and join the community at


  • Astro Teller, Captain of Moonshots (Google)



Does where you work or study support you thinking in a “moonshot” way? Are you given the freedom to be as weird as necessary to make something happen? Historically, these situations have arisen for warfare. Why do we have to wait for a war for moonshot thinking?

  • Pick a global problem
  • Propose a product/solution that makes that problem go away (the science-fiction)
  • Articulate a reasonable description of a path to building the product/solution (the science)

When you try to do something RADICALLY hard you approach the problem differently than you would when attempting to solve incremental problems.

Getting the “a-ha” takes a perspective shift. Perspective shifting is so much more powerful than being smart. Improvements are “smartness contests”. When you attempt things 10-times bigger, you have no hope of getting that done with current or traditional methods. You have to shift perspective. If failure doesn’t happen at least half of the time, we’re not shooting big enough. Do people around you “get that”?

Failures are the moments you get the acceleration you need to push on to success. Takes admitting you are wrong (humility) to take advantage of the failure. You have to humble on your way to audatiousness. If you’re moving slowly enough that you never break things or fail, you’re never going to reach the radically fast, the radically different. Orders of magnitude force you to moonshot think.


Most of the work so far has been about changing the algorithms, but what if it’s really a scale issue? By going upward 4 orders of mag. toward a neurally-inspired architecture, can we get closer to AI?

Domains: recognition (visual, speech)

Spun off into Google Knowledge


What would I work on if I knew ahead of time that I wouldn’t fail? Why wouldn’t I start on that tomorrow?