Media Dojo

Archive for February, 2011

Given the latest results from the Watson’s performance on Jeopardy last night, it would seem that strong AI claims that computers can master natural language are being fulfilled at last.

Therefore, congratulations are due IBM for enabling a machine to rise above imbecility toward having a decent shot at becoming a moron.

Watson’s mastery of natural language problems arises from a very impressive mix of hardware and software. According to HPCwire, Watson is comprised of 90 Power 750 servers, 16 TB of memory and 4 TB of disk storage. The 360 Power7 chips that make up Watson’s brain represent IBM’s best microprocessor technology. Each Power7 is capable of over 500 GB/second of aggregate bandwidth and Watson has 2,880 cores at its disposal.

But the real fuel for this Ferrari comes in the form of a software system called DeepQA, which analyzes all the colloquialisms and relationships among phrases and facts that make up a Jeopardy game. It does so by dynamically correlating in real-time all of the “as-is” information sources out there (dictionaries, encyclopedias, books, thesaurus, newspaper clips etc.) that can be used to statistically analyze a Jeopardy problem.

I don’t want to cast aspersions on IBM for this very impressive technical feat. IBM wants to sell Power 750 servers into the enterprise analytics market. Jeopardy is probably the best natural language product placement opportunity out there. So kudos for shrewd marketing.

But just as we have no issue with a bulldozer or steam shovel besting our ability to dig ditches, we should avoid either paranoia or far worse, ecstatic mysticism over the fact that Watson is probably the best Jeopardy player that ever existed.

Therein lies why Watson is effectively a moron—albeit a very smart one. The Cretan Epimenides’ timeless statement that “all Cretans are liars and this statement is true” introduced us to the fact that you can have a correct statement in natural language that is also a logical paradox. Kurt Godel took it further in mathematics with his Incompleteness Theorem, which said “all consistent axiomatic formulations of number theory include undecidable propositions”. Translated, Godel means that you can make mathematic statements (what is stated and how it’s proved) that can be complete (covers all known phenomena) but will have inconsistencies (exceptions). Or you can make mathematic statements that are logically consistent (no exceptions) but are not complete (only a small slice of phenomena). But you cannot have BOTH completeness and consistency.

Watson wins at Jeopardy because according to the rules that define a legitimate English language Jeopardy proposition (logical consistency), the machine’s algorithms and processing power will now beat human opponents—probably time and again. I don’t see a John Henry of Jeopardy out there who will beat Watson, then lay down his buzzer and die.

But I also believe that civilization will somehow endure. I’m equally confident that Watson will become a very useful tool for the enterprise analytics market as well as other logically consistent, small universes.

That said, we currently have strong AI proponents like Ray Kurzweil out there doing movie tours and selling the idea of a digital Rapture for true believing Singularitarians. The drum being pounded talks of exponential growth of processing cycles somehow leading to a direct symbiosis between man and machine to the point of being able to upload one’s mind into a database. Watson’s Jeopardy performance will no doubt be cited as a quickening step toward the fabled Singularity. A news flash Ray: given the fact that Watson is about as socially aware as a mollusk, you’ll die of old age long before biology makes true communion with silicon. If Watson’s heirs have any sense, they’ll view a lot of those uploaded human minds as pathogenic and take appropriate action.

Because of the reality and imminency of death, human beings make Art. The late poet Cid Corman taught me that most misunderstandings of Shakespeare arise from trying to impose “truths” or “meaning” where none were intended. Shakespeare’s work does NOT offer solutions to our problems, It offers us realizations of the human predicament because at certain levels, we are ALL Hamlets and Lears and Macbeths and Othellos, Antonies and Cleopatras, Timons and Troiluses. We connect with these characters because we have shards of them within us. We go to the theatre, the cinema, the concert or we go online to feel ourselves made better by experiencing consistent though incomplete realizations (thanks Kurt) of stories which fire us to explore what it means to be alive and human.

Intelligence isn’t spitting out a solution to a fucking Jeopardy problem. It’s about how you make meaning out of the limited time you got.

In media markets dominated by physical formats, channels and devices, insurgents must pull off the perfect heist.

The incumbents have not only capital and distribution muscle at their disposal, they’ve got a more deadly weapon which is time to analyze and react to competitive thrusts. One flaw in the insurgent’s plan or execution and the full weight of the incumbent’s advantages comes crashing down on rebel’s head.

When you flip that equation to digital media, incumbents must craft the perfect defense.

An insurgent has all the time in the world to find that one flaw to exploit. In the information security world, there’s an old saw that a software patch is the best advertisement for where an attacker will find their next meal. The gap between the release of a security patch and when the patch is universally implemented gives the attacker a precise road map for picking off weak or non-compliant users who don’t update their machines.

When I view how most of today’s incumbents are responding to the tidal shift of digital media from physical to virtual, from ownership to rental, from consumption to participation; I’m not seeing strategy.

I’m seeing patches.

Hollywood’s mass hysteria over streaming rentals as epitomized by Netflix isn’t as knee jerk as it sounds.

How would you like the prospect of reorganizing your business around $2 rentals as opposed to $20 physical DVD sales?

Effectively speaking, broadband and the rental model has started strangling the product sell-through business of video. Netflix is simply the current poster child of the fact that technology and some well timed licensing deals have created a more convenient model for consumers to access high-end paid video.

And as we know, convenience trumps EVERYTHING.

But let’s not get ahead of ourselves. During the 80s, the original video rental market started killing the sell-through market for VHS tapes. When the first VHS tapes for purchase came on the market, you were talking $39.95 or even $49.95 price points. So once Blockbuster started renting out tapes, people stopped buying because rental made so much more sense.

The studio response? Flatten the price curve. Drop the price for VHS to squeeze Blockbuster or force it to raise the rental fees. Tapes went from nearly $50 to more like $20. Then DVDs emerge to enable the studios to bundle in more content in a way to maintain prices for a while. And once DVD prices began their descent, physical rental became a secondary market and physical product sell-through became the big market.

Sure, Blockbuster and other physical rental firms like Hollywood Video made mistakes that hastened their demise. But it was home entertainment sell-through via cheap DVDs that killed them. People said , “screw it, I’ll buy the DVD at Walmart rather than troop down to the store”. Note, I’m not discounting Netflix’ DVD-by-mail launch model in the least. But let’s face it, if Netflix hadn’t gone the streaming route, we would start tuning up for its swan song rather than heaping praise on Reed Hastings.

Now Hollywood and the CATV industry are screaming bloody murder because studios like Starz and Epix cut deals with Netflix that offer literally digital pennies to cable dollars. Epix licenses content to the CATV industry at roughly $2 per sub per month according to some sources I know. The relative fee Netflix paid was on the order of $0.15 per sub per month. “Why the hell did you do that deal?!” bellow the studios while the cable guys are screaming just as loud “why the hell did you do that deal?!”.

“Stimulating the market” for paid streaming content may have been Epix’ motivation for cutting such a deal. However, the reality is that these deals have moved the price point down for paid video. Just like digital music prices must orbit around $0.99 plus or minus depending on features and bundling, so too are Netflix and its competitors ensuring that John & Jane Q Public believe that $7.99 per month is about right for paid streaming video.

Sure, there’s going to be a lot of words traded as the deal comes up for renewal. Netflix says it’s a friend of content providers. But they may not see things that way. In addition, the cable guys are re-evaluating what they’ll pay studios for content, especially reruns, which has a higher proportion of cream. Turner Broadcasting cited overexposure on digital platforms as one reason for not picking up “Modern Family”. True, it could have been a negotiating ploy but there’s also a grain of truth.

Thus, the battle lines are drawn. It’s a struggle by the studios to maintain physical/digital/some-kind of sell-through against streaming rental.

That’s probably why we’re going to see a lot of cloud-based digital lockers for storing purchased movie content in the coming months. The studios like the cloud because they’re betting that people won’t want to fill up their hard disk but will still want to own something. That’s Ultraviolet in a nutshell.

In the meantime, Netflix is habituating people to rental.

So what do you think? Will we look back at the original streaming deals cut by Netflix as the video equivalent of Bill Gates licensing QDOS (quick & dirty operating system) from IBM?