Wednesday, June 11, 2014

IBM and NSA visions of our cyberfuture

I just got back from the big meeting on our cyberfuture, for US government IT people, put on by IBM and Federal Computer Week. They did give official IBM views, lots of them -- but for NSA it was NOT official. It was just an hour talk by Keith Alexander, recent Director of NSA. In
listening to them, I naturally connected to other sources on the same issues that have come to me recently. (You wouldn't believe some of the impossible things I sometimes find on my desk.
I remember back in graduate school, when I set up the Harvard Committee for a Space Economy, and was trying to think of words to write about the subject, when I glanced at my near-empty trash can and found a mark up of a TV piece by Walter Chronkhite on that subject, original draft and markups.. no idea how it got there. Also, in the summer of 1971, when I found a recent classified report on Vietnam, from the US command office in Vietnam, in my desk drawer. How did that get there?)

Anne Altman, head of IBM federal services (a big one), finished the morning by talking about SoftLayer, IBM's effort to become The Cloud Provider, which they see as one of the keys to the future. They have built 40 new super cloud server farms since they acquired SoftLayer about a year ago, two of them dedicated to the US Federal government. Picture the entire US government being managed from just two relatively secure server farms, one in Dallas and one in Ashburn Virginia, the former starting up this June, and the other a little later. It sounds as if they hope they will beat out everyone else
by offering folks like government agencies a chance to "own a condo" (or rent a villa?) inside their own
complex, where the agencies will control their own space within the complex, unlike other cloud farms where people only own virtual territory, not physical territory.

Dr.  John E. Kelly, senior VP and Director of Research for IBM (worldwide) also spoke for most of an hour.  (There were so many neat well-crafted slides I wondered whether they will be on the web.)
What was most striking about his talk was the heavy emphasis on using technologies which my little program at NSF has been leading, from the viewpoint of the fundamental enabling mathematics, without which it can easily become BS. But there was a difference between current efforts and R&D directions, important to keep in mind.

Kelly had a $6 billion per year R&D budget, many times larger than any other US budget I know about for the same areas (though I have no idea what google and Microsoft fund in the same spaces).

The major near-term push is for the 'Watson" project, managed a bit like the way the initial
project to develop the IBM PC was managed. Will the whole world be run by Watson? It sounds as if they are far along in trying to make it so, with the US government only one of about 20 branch offices.
They showed a clip of Watson Debate, a very neat idea, illustrating how the system can come up with evidence on BOTH sides of a debate, neatly finessing the very fundamental issue of
differences in trust. But though it had access to all of wikipedia and more, the arguments it came up with were disturbing first-layer, superficial arguments -- both an offer and a warning at the same time.
It was neat to hear how they can use the system to diagnose the dynamics of unique specific brain tumors (a joint project with Sloan-Kettering), somehow mapping data on the two relevant DNA sequences (the tumor and the normal DNA of the person) into pathways the tumor uses to live and grow, and search tousands of medical journal articles to propose and evaluate unique chemical treatments for a specific person to kill just that tumor as safely as posisble.

For the grand future, the R&D strand, he stressed four key technologies for the real future -- bio-inspired computing, artificial synapses (neuromorphic hardware), quantum computing and graphene -- all areas we fund, at the very cutting edge of the world! (I commented to the guy next to me:
"The devil is in the details, but so are we.")  

Kelley also stressed how huge growing volumes of data from legacy systems, and inputs from new cyber collaboration, can come together to yield insight. He argued that "cognitive computing"
(depicted by a brain with a chip inside) is simply the only way possible to keep up with such a huge new flow of data, from sources like sensors, internet traffic, enterprise data and one other...

Alexander began by saying "My friend John here, whom I have worked with a lot, just said about half of what I meant to say, so I can focus on some further and related points."

He said that the future is the cloud, and has to be. He told a story of how OSD one day found 50,000 pieces of malware on their most important computer, and then changed policy by letting NSA come in and clean up, which they did within 24 hours. He talked about the new Cybercommand, already up but aiming at full force by 2015 or 2016, reminding me of Clancy's Threat Vector (as did some of the other discussion)  "We will protcet the financial centers, not just DOD itself. We protect the nation. But we also defend liberty, and work harder than you know to prevent abuse. Let me give some details about how.." He even introduced mobile phones into NSA proper...

But just this moment a friend knocked on my door, and we head off to a UN type meeting about cyberdialogiue for eatyh as a whole. (By the way, Alexander did discuss tings we could look for constrictive dialogue with with Cghina, where the issue is erspionage not attacks.. and the issue is not espionage inngeneral, but stealing problems.)