Is Speech Recognition Ever NOT Going to be Five Years Out

From Ina Fied at CNET news.com's Microsoft Blog:

Microsoft attempted to show financial analysts on Thursday that speech recognition was finally ready for prime time...Things went from bad to worse as the software got few of his words and commands correct. The closest it came was recognizing "I hope you like the DVD" as "I hope you let the DVD..." The crowd finally did make some noise, breaking into snickers as the demo progressed.

I always find it remarkable just how intractable the speech recognition problem has remained outside of very specific, narrow domains with very limited vocabularies. I think it fair to say that few people from decades past would have envisioned the truly remarkable pace of advance that computer hardware has made--and yet be unable to understand a simple spoken sentence.

Bandwidth Isn't Free

Remember how bandwidth was going to be free? Or as the subtitle of George Gilder's breathless Telecosm put it: "how infinite bandwidth will revolutionize our world." (I note that used copies of this particular work of fiction go for a penny on Amazon--but I digress.) Of course, bandwidth is neither free nor infinite. That reality carries with it implications that we see daily. Here are just a few recent examples.

Netflix. John Paczkowski over at Good Morning Silicon Valley turns his pen to Netflix' "vaporware video-on-demand service that's poised to become the Duke Nukem Forever of the industry." He quotes Netflix' CEO Reed Hastings' as saying "It would be all too easy to conclude that movie downloading was exploding in growth. The reality is the current Internet movie delivery services continue to show no growth in traffic. Nada. ... the whole industry is held back by the exclusive windows and the Internet to the TV issue. Movie downloading will evolve over the next decade, but it will do so slowly." I get the feeling from John's tone that he doesn't really buy Netflix' view on this, but I certainly don't get the sense that VoD has hit the big time anywhere.

Why? Well, I'd say it was because of Netflix, its competitors, and even the pricing changes it has forced on bricks-and-mortar rentals.As the old saying goes, you can't beat the bandwidth of a station wagon full of backup tapes and the same logic applies to sending DVDs through the mail. Plus you can easily play them on the DVD player that's already attached to your television (or on your computer if you prefer) without going through any lengthy downloads or dealing with expiring copies.

YouTube. To be sure, the issues around VoD aren't solely (or perhaps even primarily) related to bandwidth availability and cost. So let's consider the case of YouTube which is merrily enjoying its 15 minutes of Internet fame as the online repository of all sorts of videos that people upload--not all of which they actually own the rights to. The copyright issue is going to be one YouTube headache (as discussed here by John Battelle). A lot of the infringing may be nominally "harmless" to the copyright owners (such as footage from the World Cup and so forth)--either straight video snips or remixes based on them. (I briefly discussed "remix culture" earlier in this post.) But that's a topic for another day. What caught my eye more recently was this James Robertson post which mentions that YouTube's annual bandwidth cost is looking to be about $18 million.

And it's not a new issue. For example, AdCritic used to be a great site for checking out interesting television ads. Yes, that may seem a bit perverse, but the best ones are really pretty creative if you can watch only the good ones on your schedule. However, AdCritic had to go to a pay model--and a not inexpensive pay model ($100) that essentially takes it out of the casual-use realm. A major reason I suspect is that it's hard to support the costs of video streaming with ads.

Bandwidth may be more plentiful and available than it once was. But it's a long way from free. At the very least, it appears as if the ubiquitous Google ads-funded mode of supporting websites doesn't bring in enough money to cover the bandwidth costs of video streaming.

Early Reactions to Adobe Lightroom

Adobe Lightroom is a new program currently in public beta. The Mac beta has been out for a while but the first Windows version (beta 3) just came out. Essentially Lightroom combines a digital image cataloging program with a "develop" module for RAW and other image formats that can be used to apply a variety of global image corrections. (In other words, it can do things like change brightness/contrast/curves, hue, saturation, and the like but not do local touch-ups or make changes only to a selected area of a photo.) It can also create slideshows (which can be exported into formats such as Flash) and print photos individually or in groups. In at least some respects, it competes with Apple's troubled Aperture program. It also replaces the RawShooter software from Pixmantec, a company acquired by Adobe last month.

Rainy weather gave me the opportunity to spend some time with the program last weekend. Initial reaction? Intrigued but cautious. Some observations and reactions.

• The Windows version is still clearly beta--perhaps even early beta in some aspects. Stability was reasonable enough for me, but it was almost unusably slow. You really need to be able to zip through photos in a cataloging application and you can't really do that with Lightroom as things stand. From browsing through Adobe's Lightroom forums, this problem seems to be both widely observed and acknowledged by Adobe. There's no reason to believe that it won't get better, but is--at the least--something to keep an eye on with future beta releases.
• More fundamentally architectural is Lightroom's organizational metaphor. Lightroom categorizes imported photos by shoot and collection. Each photo is in one shoot only; it can be in multiple collections. Although it (optionally) allows photos to remain in their original folder locations on a hard drive, the actual hard disk file folders do not play into how Lightroom organizes the catalog. I'm still thinking about what this would mean for my workflow; currently my downloading application (Breeze Systems' Downloader Pro) names my files based on a "job code" and puts them in a directory named for the month the photos were taken. Given that one often accesses digital photos using a variety of applications, I'm not sure of the wisdom in not integrating with the physical directory where files are stored. It seems to add a level of abstraction that is neither necessary nor desirable.
• In any event, I would probably have to reorganize my files somewhat. Today, I often keep various converted versions of files and the like underneath their original import directories. I would probably want to move these somewhere else so my catalog isn't filled with multiple versions of the same photo--but this is something that would apply to just about any image cataloging program. (I currently use Breeze Systems' BreezeBrowser Pro which, contra Lightroom, explicitly works within the existing directory structure. It's a very nice program but, as the name implies, is more of an image browser than a cataloging program per se.)
• A number of the pros on the forums seem to be unhappy about the basic concept of having to pull your photos into a catalog to work with them (unlike, I gather, RawShooter). The basic issue is that some of these guys might take thousands of photos at a shoot and they want to quickly process them and get them off their systems as quickly as possible. From their perspective, Lightroom is more oriented toward more fussing with more modest scales of shooting. Although none of this concerns me directly, I see the point that this could perhaps be considered an application more for prosumers than working pros--or at least pros cranking out huge volumes of studio or other such work.
• Unlike Photoshop, the application interface is "monolithic." I don't have any particular problem with the UI overall, but I dislike that I can't drag the various control panels over to my secondary monitor like I can on Photoshop--leaving the main monitor for just the photo I'm working with.
• In general, the "develop" module where you apply corrections to the photo seems pretty complete and nice. One interesting component is the "curves" adjustment which I still find a bit of a black art in Photoshop. Adobe's made using it quite a bit more intuitive without sacrificing much power. (And, frankly, if you need more fine-grained curve tweaking than the Develop module provides, you're probably working on the image in Photoshop anyway.)

So, besides performance, my biggest issue is Lightroom's decision to basically ignore the directory structure of images on the hard drive which, at the least, means that I'll have to rethink workflows and that--I'm pretty sure--will make it less straightforward to find and deal with digital images using other software. It will also be interesting to see how this integrates with Photoshop over time--a program that has gone through a couple iterations of built-in image management (currently Adobe Bridge) and which will have to deal with some seriously conflicting requirements to both better integrate with Lightroom while also keeping its huge base of existing users content.