Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : PC Sector Round Table -- Ignore unavailable to you. Want to Upgrade?

To: Mark Oliver who wrote (1017)	10/13/1998 12:08:00 AM
From: Yogi - Paul	Read Replies (2) \| Respond to of 2025

Mark, You must be a squirrel. <<With video, you'll have many events that you'll want to keep for later viewing, >> If you assume some advance in available bandwidth, why would you store a video presentation locally when you can access it at will on someone else's storage? <<much like you keep old e-mails to recall useful conversations>> You get better e-mails than I. <<So, perhaps next year, many annual meetings will be delivered by video. New product announcements may come via video. Even quarterly conference calls could involve video. How many of these can you imagine storing and playing back later.>> None. Give me the hard copy now and the SEC filing (without the spin) later. <<Also, if you find streaming quality is lacking, think of how perfect it would be to store it and play it directly from your disk drive. >> Why the disk drive? If you must have a copy available, put it on a zip drive or an orb drive, or a 120 drive, etc, etc. <<Only last year I read 75 million IBM 3270 terminals were still in use. >> You see potential upgrades. I wonder if technology capability has passed functional necessity. And finally-- <<If you have tons of disk space, you'll use it.>> Only if it is cheap space. If disk space is dear to you, you find many things you don't really need or use. Frankly, I believe we are going to see smaller but faster disk drives and larger removable, portable storage devices at the desktop level. But then, I'm a nut, Yogi

To: Mark Oliver who wrote (1017)	10/13/1998 7:48:00 AM
From: Pierre-X	Respond to of 2025

Re: Videomail as the salvation; offline speech recognition; Labor lag. (LONG) The idea that storing video locally was kind of a spur-of-the-moment idea I had when we were all at diskcon. Let's think that concept through in more depth. Causes: Users generate video themselves, and they save videos from other sources. IF we continue to see 1. increases in density and 2. concurrent declines in cost per unit of storage along with 3. associated increases in local access bandwidth such as UDMA/66 and FC-AL, THEN extrapolating along the last two year's growth path should put us under 1 cent per megabyte before the end of 1999. Considering a typical 10 minute video stream at Net-transfer resolutions are only a few megabytes, it literally will cost pennies per videoclip to keep them around. Why will people keep them around? Because it literally will cost more to spend time thinking about which ones to delete. As supporting evidence, consider how many co-workers you know who don't EVER delete emails or voicemails. Generally, only compulsively organized people take the time to even sort their email histories into categories, much less go through and delete the unneeded ones. I have personally found tremendous value in being able to go back over what amounts to a communication audit trail between myself and various clients to clarify or retrieve information ... and sometimes to correctly place blame for failure. The storage cost for saving every email I've ever gotten, and I get 20+ emails per day: negligible. >> Why this scenario is unlikely at the moment << Videomessages have a much shorter "half-life" of usefulness for two reasons: 1. "Scanning" a videomessage or audiomessage is much less efficient than "scanning" a textmessage, and 2. Currently we have no real technology for searching against a DB of video/audiomessages. The combination of those two things depreciates the value of stored A/V messages rather rapidly. There's really no way around problem #1 but problem #2 can be attacked via “metatags” – text information attached to a particular A/V message which describes the contents of that message in some useful way. We already use metatags in our normal web browsing activities -- any time you use a search engine, you're querying against content which may include A/V materials, but the search engine examines the metatags embedded in the HTML documents which (ideally) accurately describes the contents of the document. I'm tempted to digress here about the evils of metatag abuse and how such abuse queers the wicket for everybody else, but I'll restrain myself! <g> One method of generating “good” metatag descriptors is -- my favorite topic! – SPEECH RECOGNITION. Theoretically one could sic an SR engine on a piece of A/V material to have it automatically transcribed, and add that transcription to the associated searchable metatags. (I call this process offline SR transcription, or OSRT.) Really, really powerful stuff. Imagine being able to search against an A/V DB, for example “what your country can do” and as part of your query results you get a handle to an ACTUAL VIDEO CLIP of Kennedy giving that famous speech. Imagine being able to search against your voicemail histories this way! Imagine being able to automatically ORGANIZE and CATEGORIZE A/V materials this way. Imagine the power of SORTING and JOINING A/V materials this way. The possibilities are mind boggling. As a rebuttal to the technology naysayers out there, an SR engine tasked with transcribing recorded sequences has, in one respect, an easier job than an SR dictation engine because the transcriber does not have to function in real time. Free from the realtime constraint, an SR transcriber can employ “oversampling” to increase fidelity and also has more time to access larger context databases and vocabularies, which those of you who have some familiarity with SR technology will understand. Now let me turn around and say why I think this may not happen very soon. The obvious thing is that the OSRT packages don't exist yet to create these transcripts and metadescriptors. There's also the problem of creating metadescriptors for A/V clips that contain little or no speech ... Furthermore, I think we may see a slowdown in the rate of unit storage price declines for the simple reason that capacity expansion is slowing. Unit storage prices are a function of both technology advance and supply and demand relationships, and I think these variables will move in opposite directions. I have a theory that I formulated which predicts technology advance in this sector will not slow in the near future. A: Technology advance is a function of intellectual investment by specialized researchers. B: The supply of specialized researchers changes very slowly. It takes a long time for an engineer to develop the kind of expertise needed to make fundamental contributions to the stock of human magnetic storage technology knowledge. Conversely, the incredible specialization required reduces the relative utility of these personnel in other industries or even other segments of the HDD industry. C: There was a boom in the HDD industry during the roughly period 1992 through 1996, which resulted in attracting talent to the pool of specialized researchers. ERGO: Conclusion 1: During that boom period we saw some poor products come out e.g. WD's faulty 1.6G Caviar drive which was recalled due to the green R&D personnel from (C) and the fact that there was a big push to get the drives built. Conclusion 2: We now have a large pool of incredibly expert researchers fighting for sheer corporate survival. There is NO room for fat in this business – all the marginal people will be idled or axed. Basically we have some of the worlds most brilliant minds, with years and years of experience, working to bring us incredibly dense hard disks for incredibly low prices. My guess is that we'll be the beneficiaries of some of the best products built in the history of mankind. I bought a new IBM Deskstar 14GXP series 10GB disk a couple of days ago, my fourth IBM disk in the last 6 months. These are great, great disks. But the thing is, you can basically go out and buy ANY large disk -- WD, or Maxtor, or whoever -- and be certain you are getting an amazing piece of technology. I think we'll see this go on for quite a while. Next time I want to talk a little bit about the differences between human SR and machine SR, which has to do with “knowledge pyramids” and “symbol chains”. <g> God bless, especially if you read this whole thing! <g> PX