Google Search Will Be Your Next Brain

New submitter Steven Levy writes with "a deep dive into Google's AI effort," part of a multi-part series at Medium. In 2006, Geoffrey Hinton made a breakthrough in neural nets that launched Deep Learning. Google is all-in, hiring Hinton, having its ace scientist Jeff Dean build the Google Brain, and buying the neuroscience-based general AI company DeepMind for $400 million. Here's how the push for scary-smart search worked, from mouths of the key subjects. The other parts of the series are worth reading, too.

Comments

Need better personal/collective info tools to cope

[Paul Fernhout]

http://pcast.ideascale.com/a/d...
"This suggestion is about how civilians could benefit by have access to the sorts of "sensemaking" tools the intelligence community (as well as corporations) aspire to have, in order to design more joyful, secure, and healthy civilian communities (including through creating a more sustainable and resilient open manufacturing infrastructure for such communities)."

Even just to cope with the implications of what Google is doing in AI... Still working on them, slowly...

My feeling is that our trajectory coming out of any AI singularity will have a lot to do with out moral and social trajectory going into one. So, we should do all we can now to make the world a better place for everyone, to hopefully improve that outcome.

I used to do AI in the 1980s, with my undergrad work at Princeton related to the Pointrel system maybe helping a bit to inspire Wordnet (started by my undergrad advisor as I was graduating), and (accidentally) making probably the world's first simulation of self-replicating cannibalistic robots... But in hanging around CMU's Robotics Institute in the mid 1980s, I got the disturbing feeling that it might be too easy to make "Mind Children" good enough to destroy us humans, but not good enough to "replace" us. After all, an aggressive enough self-replicating robotic cockroach could probably do in the human species, and that does not take much intelligence. As I said at a talk I gave at a conference on AI and Simulation, it is very easy to make AI and robots that are destructive (as I learned unexpectedly from my own simulations); it is much harder to make robots that are cooperative (either with each other or humans). Someone from DARPA literally patted me on the back after that talk and said "keep up the good work" -- which gave me a lot of pause, but I'm not sure which aspect he emphasized (the destructive or constructive). But that sort-of cemented my feelings, and I have not worked much on "AI" since (in an independent AI sense; one might argue any knowledge management stuff has a flavor of AI, including my Pointrel system work).

Still, as with any arms race, and that is what the current push to AI has become, and arms race whether in commercial or military terms, it can be hard to figure out some way out of it before total destruction. So, better sensemaking tools might help with that. There are other problems we wrestle with as well that they could help with, like human health issues. Such tools, as they get smarter, will hopefully be designed as cooperative platforms, for each interaction between the machine and a person, and between people, and between machines.
http://www.shareintl.org/archi...
"These words written [praising competition] by American college students capture a sentiment that runs through the heart of the USA and appears to be spreading throughout the world. To these students, competition is not simply something one does, it is the very essence of existence. When asked to imagine a world without competition, they can foresee only rising prices, declining productivity and a general collapse of the moral order. Some truly believe we would cease to exist were it not for competition. Alfie Kohn, author of No contest: the case against competition, disagrees completely. He argues that competition is essentially detrimental to every important aspect of human experience; our relationships, self-esteem, enjoyment of leisure, and even productivity would all be improved if we were to break out of the pattern of relentless competition. Far from being idealistic speculation, his position is anchored in hundreds of research studies and careful analysis of the primary domains of competitive interaction. For those who see themselves assisting in a transition to a less competitive world, Kohn's book will be an invaluable resource."

In gen