I thought I’d try liveblogging from hossein eslambolchi’s talk today at Harvard. Maybe it’ll force me to take good notes, or maybe it’ll be a mess. :) Attendance is good; apparently they are giving away five Shuffles at the end. At least one audience member running Ubuntu.
Mr. Eslambolchi is the AT&T CTO and CIO, and appears to have responsibility for the AT&T Labs, which is still pretty cool even if they are sort of a shadow of their former greatness. He is writing a book on what he thinks the tech world will look like in 2020.
Apparently his talk today will touch on (among other things) the end-to-end nature of the internet, and whether or not that is still a good idea, many years in. [That didn’t really come up much, at least until Q&A. Probably more on that this afternoon.]
His introduction is pretty funny, inasmuch as he clearly finds (or wants to convey the impression) that Harvard is a peer to AT&T- a huge, globally influential institution of the highest rank. The humor I found in this is just that Harvard is (in theory) not at all in the same business as AT&T is.
state of the telcom industry and AT&T: in 2001: overcapacity- too much investment; fraudulent players (worldcom, global crossing, enron) leading to competitive pressures to keep up with fake numbers; regulatory uncertainty after telcom act of 1996. Impacts: pricing pressure was very high- not something AT&T was used to. Bankruptcies- lots of companies stopped paying bills after the bubble. Competitive Technologies also a problem- AT&T perceived as stodgy Ma Bell company, and was, to a certain extent. had to change technological approach to compete. All these factors created perfect storm in AT&T and telcom in late 2001.
top ten tech trends he sees:
- 10. home LANs will proliferate- everyone will have ethernet/802.11 everywhere. Bandwidth demands will never diminish- only trend there is up. 3Mbit current average of pipe to home; will go to 40Mbit by 2010 and 1 Gbit by 2020. Thinks ethernet will be the connector for home stereos/TV/etc.
- 9. knowledge mining will transform the way we do business. distinguishes data or information mining from knowledge mining- thinks this will be the trend by the end of the decade. sees the mining happening in realtime- round trip to the warehouse is too long.
- 8. Wired and wireless will converge, ‘accelerating virtualization’. 2.1B cellphones by 2008. Thinks wired communication will be basically dead by 2020.
- 7. broadband will be common- causing the death of locality. broadband growth is exponential, basically- 1/2B people will be on broadband by 2015. Once everything is IP-based, won’t matter where in the hell you are (unlike area code on phones)
- 6. e-collaboration will dominate the workplace- next generation speech recognition. Universal remote control is not something that works for many of our tools- we can’t glue together email/IM/etc. the way we glue together our TV/VCR/DVD/etc. Thinks biometric security and voice recognition will move that forward. (claims voice recognition is improving 5x faster than moore’s law, which frankly sounds like bullshit.)
- 5. Sensor networks will be everywhere, esp. RFID. Of course this means IPV6. hehe- VIN (vehicle id number that is unique for every car) should be an IP address.
- 4. wireless internet will be big, which will drive mobility (yes, this is as redundant to other points as it sounds)
- 3. ‘network will be the computer’- communications and applications will converge. [This reminds me that I’d kill to have a way to fund some next-gen experimentation in conversation-centric tools/UI.]
- 2. security is critical- security is a global problem; no one takes it seriously. More madeup-sounding numbers: last year $13.5B in loss to data theft; $17B predicted this year. All students must be trained in security; must be industry-wide, not just dealt with by cisco/ms/whoever.
- 1. IP (internet protocol, not intellectual property) will eat everything. [It pretty much already has, no?][later he said ‘yes, it already has, this is not really news.’]
So, what is coming?
today in 2005:
- IP: collaboration, storage, grid networks
- Devices: flexible display, ‘perpendicular’ storage
- Wireless: mobile video, over-the-air programming, location-aware services, interactive video, wireless VOIP, seamless mobility
- Context: semantic web, ‘composite’ applications, video search, biometrics
by 2010:
- IP: self healing networks
- devices: 3-d printing, fuel cells, huge but mobile storage
- context: ad-hoc sensory mesh networks, speech-to-speech translation
- wireless: smart radio
by 2015:
- ip: wearable networks, intelligent network optical chips, private on-demand bandwidth
- devices: nanocomputers
by 2020:
- wireless: land-lines all retired
- devices: holographic storage
- context: holographic teleconferencing, tele-immersion
what is coming in the network:
- yesterday: pipes and ports; IP future: rich application-centric services. Changes business model- won’t be enough to sell bandwidth
- yesterday: individual networks; IP future: converged collaborative network and device/protocol agnostic
- y: fixed capacity- buy as much bandwidth as you might need, ever; future: on-demand utility (both for bandwidth and computing)
- y: IPv4; future: IPv6, multicast, unilink, VPLS (US military already demanding v6). says current techs are limited by IPv4 and current routing
- y: heterogeneity required; future: heterogenous by choice. means this in regard to networks.
- y: person-to-person communication; future: community-centric communication
AT&T clearly thinking about the totally converged ‘use whichever communication tech is most convenient’, up to and including the group collaboration ends of it- not just talking to each other, but scribbling on the same document, etc. None of this is new, exactly, but interesting that AT&T sees it in the same
Sees the telcom battles coming up around access- who can give you the best access from the most places, etc. Sees access in general- WiMax in specific- as a way for them to bypass LECs and other local folks and go directly to the customer. Cheaper to deploy than 3G phones and designed around IP from day one. Expects wimax to offer 100Mb/s in the next 5-10 years. Doesn’t say it explicitly, but seems to expect that it’ll kill the current cell phone networks.
transformations in the future: (ran out of time and started getting fast here :) AT&T passes 2.6PB data/day; everything is growing exponentially. Expects the network to become intelligent to deal with security threats, because it will be easier to put it centrally instead of at the leaves. They were able to predict slammer because of peaks in UDP traffic a few days before, because they mine the data so quickly. They are building an Internet Security News Channel- 24/7 streamed video for customers- shows, talks, tutorials, breaking news.
Impressive demo of customer interaction with a computer, answering a billing question. They have taken it live for customers- toys’r’us, for example. See it as a product more than tool for internal use, which is interesting.
Sweet demo video of real-time voice recognition, including on-the-fly learning of proper nouns. Demo was clearly ‘filmed’ in metacity- labs researcher is using GNOME, I guess. Cool :) [Eslambolchi is using Windows, of course :)
Control v. Freedom- believes industry should move towards freedom. yay :)
- control: passive reception of information and entertainment; freedom: active participation; choose when/where/how
- control: limited news providers with fixed newscasts; freedom: blogs, personal video studios- the impact of billions of webcams.
- control: proprietary solutions; freedom: standards bodies
- control: proprietary software; freedom: open source (specifically cited bugs as a driving force)
- c: licensed spectrum; f: smart radio over unlicensed spectrum- bits/hz will increase, so needs more hz
- c: regulated access; f: open access
- c: proprietary networks, vendor dependency; f: open networks, architecture, and API. Smart companies should become dynamic software companies instead of implementing things in hardware- become MS instead of IBM implied.
- c: sync; f: async
Q&A:
Is a holder of 700 patents, but thinks frivolous patents are a problem. Implies competition needs to be on actual innovation embodied in code. Totally sidetracks to P2P stuff; thinks there will continue to be battles for ages about cloaked ‘illegal’ software v. law/legal forces.
end-to-end question; thinks that it will be hard to wean tech consumers from edge-centric security, but that it is necessary because the edges can’t be as reliable/up-to-date as things in the middle, because the middle can do data collection+mining, which the edges cannot. Thinks it’ll be a decade or two before the next-gen internet will solve some issues to make the next ‘net capable of being secure while not getting in anyone’s way very much. Focus is of course on corporate networks. [I used to think Yochai Benkler’s argument that wireless will save us from corporate-commercial-regulated network monitoring was insane, but with WiMax, that might not be completely insane anymore.]
Data mining is purely on the packet headers, not on the packet contents generally because of privacy concerns.