Lecture: Metcalfe | September 25
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Subtitle |
| |
| Title of Series | ||
| Number of Parts | 13 | |
| Author | ||
| License | No Open Access License: German copyright law applies. This film may be used for your own use but it may not be distributed via the internet or passed on to external parties. | |
| Identifiers | 10.5446/65448 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
2
3
10
13
00:00
Connectivity (graph theory)Connected spaceStandard deviationComputer animationMeeting/InterviewLecture/Conference
01:21
Connectivity (graph theory)Connected spaceProcess (computing)WeightSurface of revolutionPhysical systemLecture/ConferenceComputer animation
03:13
Presentation of a groupRight angleMultiplication signModel theoryHypothesisMany-sorted logicPoint (geometry)Radical (chemistry)Degree (graph theory)Lattice (order)ResultantProcess (computing)Different (Kate Ryan album)Product (business)Applied mathematicsSocial classEvolute2 (number)Event horizonPhysical systemCharge carrierMathematicianObservational studyUniverse (mathematics)Term (mathematics)Standard deviationBand matrixComputer programmingConnected spacePosition operatorLecture/ConferenceMeeting/Interview
12:41
Scale (map)Polar coordinate systemProduct (business)2 (number)AreaLocal ringLecture/Conference
14:07
RootMathematicianWave2 (number)Connected spaceContrast (vision)Physical lawPower (physics)MathematicsNumerical analysisModel theoryMultiplication signAreaBuildingStudent's t-testMeeting/InterviewLecture/Conference
19:22
ModulformMultiplication sign1 (number)Connected spaceNumerical analysisSet theoryAdditionLecture/Conference
21:06
1 (number)2 (number)Connected spaceHypothesisLecture/Conference
21:57
Graph (mathematics)Condition numberDivisorCausalityConnected spaceGoodness of fitCross-correlationMultiplication signParameter (computer programming)2 (number)Lecture/Conference
24:19
Physical lawNumerical analysisSlide ruleSquare numberMetreConnected spaceLinearizationLecture/ConferenceMeeting/Interview
25:28
Graph (mathematics)GradientPhysical lawSlide ruleProduct (business)Student's t-testMeeting/InterviewLecture/Conference
26:11
Polar coordinate systemConnected spaceMechanism designMeeting/InterviewLecture/Conference
27:07
2 (number)Lecture/ConferenceMeeting/Interview
27:47
Charge carrierModel theory2 (number)Lecture/ConferenceMeeting/Interview
28:29
Process (computing)MathematicsDifferent (Kate Ryan album)Lecture/ConferenceMeeting/Interview
29:14
Many-sorted logicOrder (biology)Goodness of fitLecture/ConferenceMeeting/Interview
30:06
DecimalProcess (computing)Algebraic structureUniverse (mathematics)Lecture/ConferenceMeeting/Interview
30:48
Data transmissionTerm (mathematics)Meeting/Interview
31:32
2 (number)Many-sorted logicMeeting/InterviewLecture/Conference
32:18
Bounded variationMathematical modelFlow separationProjective planeGroup actionMultiplication signRandomizationDegree (graph theory)DeterminantLattice (order)Model theoryTheoryLecture/ConferenceMeeting/Interview
34:29
Projective planeProduct (business)Object (grammar)Ring (mathematics)Physical systemDifferent (Kate Ryan album)MathematicsLecture/ConferenceMeeting/Interview
36:48
AverageLecture/ConferenceMeeting/Interview
37:29
2 (number)Lecture/ConferenceMeeting/Interview
38:25
Right anglePhysical lawMultiplication signExpected valueLecture/ConferenceMeeting/Interview
39:07
Numerical analysisMultiplication signPredictabilityCollisionLecture/ConferenceMeeting/Interview
40:41
Spectrum (functional analysis)Multiplication signFunctional (mathematics)Lecture/ConferenceMeeting/Interview
42:54
MathematicsKörper <Algebra>Decision theoryFigurate numberTriangleMultiplication signCartesian coordinate systemOrder (biology)Parameter (computer programming)MeasurementEnergy levelLecture/ConferenceMeeting/Interview
44:56
Connected space2 (number)Limit (category theory)Process (computing)Lecture/ConferenceMeeting/Interview
45:35
Generating set of a groupLecture/ConferenceMeeting/Interview
46:16
Arithmetic progressionGraph (mathematics)Cross-correlationMultiplication signNuclear spaceConnected space2 (number)Lecture/ConferenceMeeting/Interview
47:07
Representation theoryKörper <Algebra>Generating set of a groupCartesian coordinate systemLecture/ConferenceMeeting/Interview
48:17
Decision theoryOrder (biology)2 (number)Distribution (mathematics)MathematicsTerm (mathematics)Lecture/ConferenceMeeting/Interview
49:35
Lecture/ConferenceComputer animation
Transcript: English(auto-generated)
00:00
My name is Anna Wienhard, I'm director at the Max Planck Institute for Mathematics in
00:30
Sciences and also the scientific chairperson of the Heidelberg Laureate Forum Foundation. And so now it's my pleasure to introduce the first laureate for the first laureate lecture this
00:41
morning and we really have one of the freshest laureates here with us this morning. Bob Metcalf received the ACM Turing Award in 2022 and he received it for the invention standardization and commercialization of Ethernet. So this technology is crucial to connect computers
01:02
and by connecting computers also to connect people. And connection and connectivity is the subject and topic of his talk so it's my pleasure to ask Bob on the stage to talk about connectivity is a thing. It is the thing. Bob the floor is yours.
01:28
Good morning. Connectivity is a thing. It's the thing. It's my advice is that's where you start in the design of systems. So I'm going to tell some you notice I'm not using powerpoint.
01:44
I was on the board of the company that produced powerpoint. We sold it to Microsoft in 1987 for 14 million dollars. So I've been not using powerpoint for a very long time. I'll warn you I'm also susceptible to the hammer syndrome. You know
02:01
if a hammer is your only tool everything looks like a nail. Warning connectivity is my hammer and everything looks like a connectivity problem. Of course and here's what connectivity is. It's the apparatus that moves stuff from A to B. Stuff like energy, mass,
02:23
or signals which is what most of us do. So I'm going to next and by the way the Chinese have of course have a saying about connectivity. They say if you want to be rich first you build a road. Road being a connectivity technology. The Romans of course are famous for their roads. All roads
02:43
lead to Rome. And then there was my forebearer Blind Jack Metcalf who built the first 180 miles of the industrial revolution out of Yorkshire, England. So apparently runs in the family. So I'm going to tell some stories from ARPANET, Aloha Network, Ethernet, and Internet.
03:06
And there are participants in the construction of those nets in the audience and their their job is to keep me honest which is a very tough thing to do. But by the way does anyone have a question?
03:21
Do you have a question, Vint? What is your question? My question is what led you to develop the Ethernet in the first place? And I assume that you're going to tell us that. I am. It's right here. This is my presentation and it's right here. And questions are welcome at any time and corrections to anything I say I do
03:47
tend to exaggerate here and there. So I was not exaggerating. I was born in 1946 so I'm among the oldest boomers and I was really lucky because of what happened in 1947. In 1947 wireless television began its commercial ramp. In 1947 the words computer and bug came
04:09
into usage. The microwave was invented and in particular the transistor was invented in 1947 and my entire life has been driven by the transistor. And it led me to invent Ethernet.
04:24
The reason I'm here is I invented Ethernet. There are several other inventors, one of whom is sitting right over there, Butler. But as a result of being the inventor of Ethernet I get to be a member of the connectivity society and we meet quarterly. Our last meeting was very
04:42
sad because one of our members had died. The inventor of USB had died. So we all living members served as his pallbearers and carried his coffin across the street to his waiting grave. And we lowered the USB inventor's coffin into the grave and it jammed. So we had to pull the coffin
05:08
out of the ground and rotate it 180 degrees and then lowered it into the grave. Most of my connectivity stories begin with the two monopolies that dominated connectivity at
05:26
the start of my stories, AT&T and IBM. And the AT&T story is best told by what my mother said. When you get to Boston, Bobby, call us to let us know that you're okay.
05:44
Let it ring three times and then hang up. My mother loved me but not that much. AT&T had a product called the 500 telephone which they sold between the years 1950 and 1984.
06:04
That's a little bit different than the iPhone pace of announcements. And that's a consequence, I believe, of the monopoly position they maintained. Ditto IBM that had at the time, at the beginning of all these stories, was selling batch processing punched card systems
06:23
and promoting a thing called system network architecture that we should use to connect all of our computers. I used to call it SNAA just to annoy them. On the other hand, in 1981 they produced the IBM PC which was a big event in the evolution of connectivity.
06:42
While all that's getting ready to happen, I went to MIT starting in 1964. I studied architecture, mathematics, physics, management and EE, getting degrees in the last two. All of those were synonyms for computer science. The term computer science barely existed so at MIT you study EE if
07:05
you wanted to study computer science and at Harvard it was called applied mathematics. So I'm a mathematician it turns out. And while at MIT I worked programming computers but it
07:20
came time to build something so I built a memory for a computer. It was a class project. The course was called 6.721. I built an acoustic delay line memory and there's a big cable and you launch acoustic waves that take a while. They go from point A to point A and then you recover them and
07:40
send them back around and that serves as a memory for a computer. And that was sort of an embodiment of the idea that if you want to teach something, if you want to learn something, the best thing to do is to teach it and the best way to teach it is to build it. And so I got to build this memory. And then as I was leaving MIT, you're going to be surprised by this. My advisor was Marvin Minsky
08:07
and I did a thesis, an undergraduate thesis for Marvin and it was on a neuron model and some of its information processing capabilities. This is 1968. Marvin got the Turing award in 1969 probably
08:24
for my thesis. And I could tell by looking at Marvin that he didn't like my thesis much. I think I think Marvin didn't think I was smart enough to do AI, so he sent me down the hall to do networking instead and it worked out. My advisor was JCR, my new advisor was JCR Licklider,
08:46
the internet visionary. Speaking of visionaries, there was a guy named Bob Taylor who graduated from the University of Texas and went to work at ARPA financing computer science research.
09:00
And they were promoting, ARPA was promoting interactive time sharing. And Bob noticed he had three terminal, the story goes, he had three terminals in his office. He says, why do I need three identical terminals to connect? We need a network here. And so he, the story goes, launched the funding of the internet because those three terminals annoyed
09:25
him in his office. And the way he did that was to fund the creation of a packet switch for connecting things together. It's called an IMP. And I got number six. So when MIT came to
09:40
get on the internet, they looked around at somebody who was practiced at taking things and sending them serially down a long wire. Because I had built this memory, this acoustic memory. And so I convinced MIT to hire me to put MIT on the internet because I had built this acoustic memory. That was a baby step towards something that would get much bigger.
10:06
Bob Taylor then became my boss at the Xerox PARC. And he gave me the networking job again, only this time it was to network a bunch of computers on every desk. Imagine that, Butler,
10:24
computer on every desk. I was the first person in the history of the world to have that problem, connect those machines together. So I immediately got help. His name was David Boggs, may he rest in peace. And he and I designed, he and I with the help of Butler and Chuck and others designed
10:44
a thing that today is called Ethernet. And we had three technologies we chose. We chose a coaxial cable with a Gerald's tap to puncture it so you could install a PC without cutting the cable. We used Manchester encoding from the University of Manchester that was on
11:07
half the time. That allowed us to do a carrier ascension, collision detection to get our packets onto the internet. And finally, we used the Aloha network for randomized retransmission. So those three technologies, Gerald, Manchester, Aloha, none of which is on Ethernet today.
11:27
They were only in that first Ethernet 50 years ago. What Ethernet really contributed, I think, is first it was packet mode. It took packets all the way to the desktop, not just
11:41
characters. Second of all, it was fast. It was 10,000 times faster than what it replaced. My desk had a 300 baud modem on it. One day, the next day, it had a three megabit per second network. 10,000 is a pretty big step up. And so after Ethernet, bandwidth became abundant
12:03
so that it makes sense to upload your cat pictures. And then finally, the Aloha network isn't really operative. The third contribution was standards. Ethernet joined the TCPIP standards movement. And then after we built these Ethernet's to connect them together in an office,
12:24
we needed an internet packet, which we called PUP, the Park Universal Packet. And built inside of Xerox, the team built an internet. And then in 1979, aiming for even greater scale,
12:41
I started a company. Now in Silicon Valley, that's not unusual. Everybody starts companies in Silicon Valley. So I was just following the trend. In 1979, founded 3Com. Our first product was consulting and then a book, and then TCPIP software, and then an Ethernet transceiver and a bunch of Ethernet controllers for various PCs. And we then joined a war.
13:07
It was called the LAN wars. Ethernet versus IBM token ring versus General Motors token bus versus a few others. And that war lasted 20 years. And we won. Ethernet won.
13:26
And now this is, while it's winning, the world is not sitting still. The platforms were changing. We were going from card processing, batch processing, to interactive terminals, to personal computers, to local area networks, to client server computing,
13:46
to smart mobile telephones and so on. So the platforms were changing that needed to be networked. And the traffic was changing. Originally, it was ASCII text. That was our triumph is to get a text, a letter to go from Boston to UCLA, back to Boston again, hooray,
14:05
in half a second. So the traffic was changing. Today, the internet is 82% video. And it was never designed to carry video, but it has somehow been modified to work. Steve Jobs played a role in my little company. He took me to lunch and I proposed to network.
14:26
He was recruiting and I was selling. And I wanted to sell him a network for his Apple IIs. I called it Orchard. That's an engineer's version of marketing. Steve returned the favor by
14:44
insisting that I buy an Apple II and run VisiCalc on it to produce the business plan for 3Com Corporation, which I did. And we raised our venture capital. And he invited me to the Toy Story premiere in Cupertino. And on the red carpet afterwards, Steve Jobs came by and I said,
15:07
Steve, I want to remind you that every pixel of your beautiful movie was carried on ethernet. And he looked back and he said, thank you. And I've been living off that ever since.
15:22
During the history of this connectivity wave that's hit us, there were some reversals that I find really interesting. For example, there was an IBM executive named Herb Grosh, who had a law, Grosh's law, which said that the cost of a computer went up as the square root of its power. And the consequence of that law was to build bigger computers. Now he worked at IBM,
15:45
so it made sense for him to build bigger computers. But then there was this other guy, Gordon Moore, who came up with semiconductor versions of small computers. And so there was Grosh touting big computers and Moore essentially touting small computers. And guess who won?
16:07
Moore won. And Silicon Valley in that moment, Silicon Valley moved from Boston to Palo Alto because of the adherence to those laws. Another law was the Negroponte switch, it's called.
16:23
It was for a long time believed that short-haul traffic in the building would be carried on AT&T copper. And long-haul would be done using wireless microwave. And Professor Negroponte noticed that what was happening was the reverse. We were using Wi-Fi, wireless Wi-Fi in the local area. And the long-haul was on optical fibers.
16:47
That was a switch. And the third reversal of note, I call Marconi Cooper. Marconi spent his life trying to get his radios to go further.
17:00
It turns out all the advance in radio in the last 100 years has been getting the radios to go shorter, to have smaller and smaller cell sizes. That switch happened about 100 years ago. So today, Ethernet for which the Turing Award was given, what is it if it's not
17:21
Gerald Manchester Aloha, what is it? And I claim it's a brand of connectivity. And a brand is a promise. So when you buy Ethernet, one promise is it's a de jure standard. In this case, the IEEE standardizes it. It's packet mode, so it's native to the Internet.
17:44
It's always built faster than anyone imagines is needed. Build it and they will come. And so far, they've come every time. Ethernet is now beginning to run at 800, listen, 800 gigabits per second. Contrast that with 2.94 megabits per second.
18:04
And it's interoperable and cheap and fierce competition among suppliers. That's amusing to watch. Dave Boggs, my co-inventor, noticed that the war, the land wars had become a war of
18:22
mathematicians. We had math that said Ethernet would work. They had models that said Ethernet would not work. And it really got ugly. Dave Boggs in front of 1,000 people said, it seems that Ethernet does not work in theory, only in practice.
18:46
So we have been overwhelmed by connectivity. And that has led to a number of pathologies. Even in 1973, I wrote a memo documenting a hack. Two high school students hacked into the Internet
19:01
for Christmas in 1973. Pornography was the next pathology. And the U.S. government almost outlawed the Internet because it was so good at carrying stuff like pornography. And of course, they passed a law called the Communications Decency Act. And so now there's no more pornography. Advertising was the next pathology. And those of us involved in the early
19:29
days of building the Internet were offended that ads were being carried on our network and sought to have them removed. The first ad, by the way, was placed, the first paid ad was placed by AT&T. What year was that? 1994, I guess. But then we discovered that advertising was not
19:50
a pathology. It was the method of funding the entire Internet. So as you know, Google and Facebook and others are all advertising supported. So that pathology went away. Today,
20:05
we have a new set of them, fake news, loneliness, other pathologies. But those pathologies, like the ones that went before, I am confident will be handled in the fullness of time. Now, we're overwhelmed with bandwidth, but there's more coming. The gigabit, we are moving now from
20:30
the megabit Internet to the gigabit Internet. So if you had too much connectivity before, you're about to have a lot more of it. And I take that as good news, but not everyone does.
20:42
And there's other forms. Connectivity isn't only the Internet. There's other forms of it. One of my favorites is the Voyager space vehicle. Have you all heard of that? Launched in 1977, two of them. One of them, number two, is now 20 billion kilometers from here. So even further than the
21:04
airport. And it's receding at 15 kilometers. And I don't know whether it's per hour or per second, but it doesn't matter. It's really far away. And that has a radio in it. And the radio used to run at 150 kilobits per second. Now it runs closer to 10 bits per second because its plutonium battery
21:26
is wearing down. And I was delighted also to read that it uses Manchester encoding. So it's just like Ethernet. Another use of connectivity is it's called artificial
21:42
intelligence. It's pretty hot. Uses neural networks like the ones that I discovered and invented in my thesis for Marvin Minsky in 1968. And suddenly all copies of that thesis have been lost. GPT-4 is reported to have 1.6 tera parameters, which are analogous to synapses.
22:12
It's not an official number. 1.6 trillion parameters. There are rumors that the next
22:21
version of the GPTs will have 100 trillion parameters. But wait, the human brain has 10 to the 11th neurons with an average fan out of 10 to the fourth synapses. So even when we get to 100 trillion in our neural nets, we're still a factor of 10 short of the connectivity that we
22:45
need to achieve human intelligence. And then there was the arrival of connectivity for COVID. COVID-19, 2019 coincidentally, was the 50th anniversary of the internet. And it was as if
23:03
we built the internet for COVID because it rose to the occasion. People who said they could never, people at the University of Texas who said they could never teach over video suddenly had to. And they were, and it worked. And so I proposed to rename the acronym COVID.
23:26
COVID-19, COVID is short for collaborative video. So I claim that the most important new fact about the human condition is that we are connected. And we're suddenly connected. When I say suddenly, we started on October 29th, 1969,
23:46
and we now have 5 billion people on the internet, about two thirds of the planet. And I claim that's sudden. And that accounts for some of the pathologies that we've encountered. Against the pathologies is the value of that network. There are graphs that show the turning down of poverty
24:06
around the time that the internet arrived. Now that's correlation, but correlation is a good start toward causation. The most important new fact is we are now suddenly connected. Now we have 22 minutes and 31 seconds if anyone wanted to ask a question. Yes, please.
24:27
What happened to Metcalfe's law?
24:43
Oh, that's here. So in, there's a law called Metcalfe's law. There's Moore's law and there's Grotius' law. There's a bunch of laws floating around. So of course there's Metcalfe's law. Metcalfe's law says that the value of a network grows as the square of the number of attachments.
25:04
That's really the number of possible connections is N squared. And it was a slide that I produced at Stanford University on an Alto that we had given to Stanford that said that the cost of one of our three com networks was linear in the number of machines you had,
25:22
but the number of possible connections grew as the square. And as you know, the square often overtakes the linear. And I made a 35 millimeter slide of this graph that was before PowerPoint. And I made six copies because that's how big my Salesforce was.
25:41
And we went out and told all of our customers, the reason your network is not useful is that it's not big enough. And the remedy to that is to buy more of our products. So that slide became Metcalfe's law and it's been attacked by professors and grad students repeatedly over 25
26:00
years. I'm defending it to the end. It's a vision thing, Metcalfe's law. Sir, Marty. First of all, thanks for a great talk, Bob. And secondly, most people probably don't know
26:22
what Alohonet is. So could you please say a little more about the connection between Ethernet and Alohonet? And if you'd mentioned my academic grandfather, Norm Abramson, I'd appreciate it. We were looking for a way to have PCs take turns on this shared cable in Ethernet.
26:46
And there are lots of ways, but most of them involve adding more wires and more connectors and stuff. And I happened to read a paper by your grandfather? Norm Abramson, who probably invited me to Hawaii. And so I went for a month. The Alohonet network is a packet radio network,
27:06
was a packet radio network in 1970. And the mechanism was quite simple. You would send your packet into this shared medium, and if you got an acknowledgement, done. If you didn't get an acknowledgement, you'd do it again. But there was the danger that there was somebody else interfering
27:25
with you, so you randomized so that you didn't run into the same interference over and over again. Randomized retransmission. So I went to Hawaii and I learned two things. One, radio is not practical for a local area network. In 1973, the radio modem was this big, and it sent data at 10
27:45
kilobits per second. The other thing I learned is that randomized retransmissions would work, and we did some simulation and modeling of it. But then since we were on coax running at 2.94 megabits per second, we modified ALOA. We just took randomized retransmissions and we added
28:05
carrier sense collision detection and back offense stuff. So I'm in a constant struggle with the University of Hawaii about how much of Ethernet did they invent, and they invented that much randomized retransmission. And it works. It's been working. Is that the answer you're looking
28:22
for? Other questions? We have 18 minutes and 23 seconds. Yes, back there. So thank you very much for the very nice talk. So my question is about your process of inventing
28:42
things and the role of interdisciplinarity. Because you said you were doing artificial intelligence research, you were creating this memory, and this led to the invention of Ethernet because you reused some things that you learned there. And today, when you look at sciences, you would see that you crossed silos because these are different subdisciplines in computer
29:03
science or mathematics. And so do we need more interdisciplinarity? Which role did it play in your process of creating the Ethernet and other things in your life? I don't think there's anyone here who would disagree with the growing advantages of multidisciplinary everything.
29:25
One person who believes this became president of Arizona State University and his first act was to abolish all the engineering departments, which I view as a step forward. Did I get your question?
29:44
Yes, you got my question. Oh, but you want an answer. No, it was also a good answer. But if you look at Europe, it's often that they do the opposite. So they create departments. I mean, they merge some departments in order to mix some sorts of sciences again. Do we need different education to open another big topic where you probably have a good
30:05
opinion on? Yes. A symptom of this problem that we're discussing is I've never met a department chair who will admit to wanting the job. The department of this or that, you talk to the
30:21
chairman and they can't wait. I'm going to do this for two years and then I'll get out of here. This is terrible bureaucracy, you get a minister view and so on. So that's sad when the leaders of these departments don't want the job. It would be nice to have somebody who wanted the job doing it. And I'm sorry, I'm not really getting your question. I can help you. You can help me?
30:44
He's asking whether or not more interdisciplinary structures in the university would enhance
31:01
the kind of innovation that you exhibited in the creation of the Ethernet, mixing the engineering technologies, radio and digital communications and the fiber, not fiber, coaxial cable. He's just asking about whether we should be mixing things together more formally by having
31:24
organizations that are not departments that are narrow, but rather organizations that are broad in terms of the disciplines that they encompass. I actually think I already answered that question. Yes, we want more multidisciplinary. I'm reminded of the professor who I encountered who gave a talk
31:40
and said his course included both kinds of engineering, mechanical and electrical, which is sort of a jaundiced view of multidisciplinary. Other questions? 16 minutes and 53 seconds. Please.
32:06
Thank you for the talk. Thank you. I wanted to know your thoughts on time-sensitive Ethernet and its use in automotive industry these days. So I mentioned the land wars, which occurred in the 80s and 90s, and Ethernet won. But one of the last battleground was
32:27
time-sensitive, or it was called determinism. Because of this randomization of retransmissions, if you wanted, you could develop a mathematical model that showed that the packet would never get through, or it wouldn't get through with probability 10 to the minus 13th.
32:45
And I just attended a meeting in São Paulo, Brazil, which explains the bags under my eyes, of the automotive Ethernet group of Project 802 of the IEEE, which was full of Germans, because apparently they all make cars. And so there were Porsches and Mercedes and so on.
33:06
Porsches and Mercedes and so on. And so there is now several, and that's a problem, several Ethernet, time-sensitive Ethernet variants. And it's deeper than time-sensitive.
33:21
There's time-sensitive, and then there's time-critical, and those give rise to their own variations. So determinism is still an issue after 50 years. We're still arguing about it as to whether it's tolerated. That is, being able to tolerate some degree of randomness in the back-off.
33:44
And I believe it was handled a long time ago by David Boggs, who said it apparently works in practice, but not in theory. But I also learned in that lesson that you can make a model do almost anything you want. And you have to be careful with mathematical models. And having
34:01
had them used against me, I'm really sensitive about it. Your network will never deliver the packets with probability 10 to the minus 14. The probability is higher that all the molecules in this room will go to that corner, but it's still not zero. May I answer your question?
34:23
Any other questions? Sir? Could you tell us why Ethernet won the land wars? Could you tell us why Ethernet won the land wars? Was it technological, business strategy,
34:40
market forces? First of all, Ethernet works. And that's an important criteria when you're selling a product to customers. It was an open standard, approved. It is. IEEE project 802.3, and the customers like that.
35:03
IBM noticed, and General Motors, noticed that Ethernet was gaining some speed at 802. So they came in with their own alternatives, token ring and token bus. But IBM's heart was not in it.
35:20
They're dark little hearts. They said it's standard, but my company shipped IBM token ring before IBM. And the problem was, I mentioned earlier SNA, system network architect, you needed to sprinkle SNA dust over your ring to make it work. So we didn't sell any.
35:42
Customers wisely decided not to buy something that wasn't supported. But they detected this lack of sincerity, and that's how Ethernet won. It worked. It was a de jure standard. It was half the price, twice the speed. Those are pluses. And we had to fix one thing. We were using
36:07
coax, and that was a killer. So we switched to twisted pair, just like this. And suddenly, that objection went away. So my little company, I learned that there's more than one kind of engineer. There's quantity one engineers. That would be me. You build one, you barely get it
36:24
to work. And then there's quantity 10 engineers, and then there's quantity 100 engineers. And we had to recruit quantity million engineers, a million a month of these Ethernet cards for four or five hundred dollars each, do the math. But we learned there are different
36:42
kinds of engineers involved. Other questions? Yes. Am I on? Okay. Thank you for the... I think you also, if I recall correctly, had a brilliant licensing scheme for your patent,
37:01
which is you granted a free license to anybody who produced an entirely conforming product. So you squeezed out people who were trying to make little variants. I don't remember that. There was a... Xerox did license the patent, the Ethernet patent for a thousand dollars
37:24
to anyone who wanted it perpetual, non-exclusive. And there were some requirements, but I forget what they were. I don't think it was this... Because I remember AT&T came to my company and said, we liked this Ethernet thing, but we want it 40% faster because we have calculated that the win in the marketplace,
37:44
we needed an advantage of 40%. Would you build us a 14 megabit per second Ethernet? And my engineers refused to do it. I couldn't make that. I said, this is AT&T, they might buy a lot. And the engineer said, no, we don't believe in that.
38:04
But one of our even more clever engineers, Ron Crane may rest in peace. He made the transceiver that would run at 10 or 14. And then AT&T never used the 14 in the megabit version. So that's to your question that relates
38:23
to making... We couldn't make AT&T go to 10. We let them go at 14. Sir, Eric, good to see you again. You famously predicted that the Internet would collapse.
38:44
It didn't, but I thought it was a reasonable expectation. Can you tell us about why you thought it would? And you can tell them how it ended, too, if you want as well. But it was an interesting time and it really was growing quickly. The case for me, the Metcalfe
39:01
law wasn't helping the N squared with N being very large. Right. So in 1995, I had some tools on my desk that told me how many packets, TCP packets, were being dropped. And the number hit 25% in 95. And I know the TCP doesn't handle lost packets,
39:22
I'm sorry, that well. So this rang bells. Uh-oh, the web is taking us by surprise. The traffic is ramping rapidly and we're starting to lose packets. Uh-oh, we're heading toward a collision. And I was a columnist at Infoworld then, and I had to come up with something interesting every Thursday night. So on one Thursday night, I wrote a prediction that the
39:44
Internet would collapse during 1996. And then my readers made me agree to eat the column, if it didn't because they were annoyed. Well, it didn't collapse during 1996. So in 1997,
40:00
I ate my column, uh, admitting that I had been wrong and they loved it. My customers loved it that I was admitting that I had been wrong. So I learned a lesson there. Admit when you're wrong. Um, so that was, and there's a book about this, it's called Internet Collapses and it's still available. The Internet apparently is still up and, uh, the book is probably still
40:25
available. Exactly. CSMA has been there since a long time. Here. Hello. I can't see you.
40:45
All the way in the back. Hello. Hi. Hi. So, uh, CSMA has been there since a long time and it's still the best way how we access unlicensed spectrum today. Um, it has been there in Wi-Fi
41:01
since the nineties. It has been a foundation of Wi-Fi since the nineties. Do you see anything better coming up than CSMA? No. Good answer. Thank you. So you bring up Wi-Fi. Uh, in 1973, we went to the University of Hawaii, which had a packet radio
41:26
network. And then we invented ethernet and people have asked over the years, why didn't you use radio like Aloha back then? And the answer I, which I've already given is there were no semiconductors that could support an economic version of that radio. But in the nineties,
41:44
those chips arrived and a product, a branch of 802, the IEEE 802 called, uh, dot 11. Its official name was wireless ethernet. And then in 1999, they changed its name to,
42:01
wisely changed its name to Wi-Fi. And so, uh, so there's CSMA, there's CSMA CD, there's CSMA CA, and they all have advantages. And I, I, I'm not qualified to answer your question now. So Wi-Fi continues to advance. It gets faster and it's adding functionality. And I call it
42:23
wireless ethernet, of course, um, so that I can have credit for having invented Wi-Fi. There's a hundred people who invented Wi-Fi, but now I like to annoy them by calling it wireless ethernet. But the reason was we had to wait 25 years for the semiconductors to arrive.
42:42
And then suddenly there was Wi-Fi. Okay. Six minutes to go. Yes.
43:00
Hello. Um, broad question for you. If you were to go back and change anything about anything, doing it all over again, what would you do differently? Um, if nothing, then what do you think the field will need to change in the future?
43:24
My life has worked out so well that I wouldn't risk changing anything in the past. Cause you never know what will happen, but to go one level deeper, uh, I would have learned how to sell sooner than I did. I, I had to learn how to sell when my company almost went bankrupt.
43:47
As a measure of our desperation, they made me head of sales. They made me head of sales. Measure of our desperation. They made me head of sales and marketing. And, uh, so for example, IBM gave me two chances to convince them to use ethernet and they paid me $2,000 each time.
44:06
And I went to Franklin lakes, New Jersey and research triangle park and pitched them on ethernet, but I didn't understand selling. So I didn't figure out what their considerations were, who the decision makers were, what the applications were. I just went in and out argued
44:21
them because I had a five year headstart and, and that's an example of winning the argument and losing the order. So I did not get the order. And I think to this day, I think if I had been a year ahead and learning how to sell that I might've succeeded and that would have saved 20 years of IBM hell, uh, with the token ring, then I failed. So my second answer to your
44:46
question is I would, uh, would have learned how to sell earlier. Yes. Okay. Three minutes and 55 seconds. You mentioned that connectivity is your sledgehammer. Um, I'm curious, is there anything
45:07
you think, like, are there limits to connectivity? Are there things that shouldn't be connected or can we be connected too much? I didn't really hear your question. I'm really sorry. Sorry. Let me, ah, there we go. You said that connectivity is your sledgehammer. So my question is,
45:25
are there limits to how connected we ought to be? Can we be connected too much or should certain entities not be connected? Uh, it is my job to argue in favor of connectivity.
45:42
I am not being fair about this. Uh, and I don't know the answer. And when you look at fake news and all those modern pathologies, they raise the question whether you can have too much. I know you can have too much too soon because we've just lived that. But can you have too much over the path in the fullness of time? I don't know how to answer that.
46:03
Think about how our generation and distribution systems should be connected to the internet. Of course. Yes. Aren't they already?
46:25
Interesting. I love this graph that I've mentioned earlier that shows the progress of poverty over time. And it's correlation of course, but there's this unprecedented downturn that occurred
46:44
following the arrival of the worldwide web. And I attribute that downturn on poverty to the connectivity that was provided to all those economies around the world. And that's a lot of good that got done there. And it's hard to balance that again. I just saw Oppenheimer,
47:01
so I'm fresh on nuclear weapons. A minute and 43 seconds. I promised to finish on time. So that's why I'm, because I believe in keeping my promises. One more question. What's that? Yes. Yeah. When
47:22
it suits me. So I guess one question I have, one question that I have is now that like these, we've seen new technology, especially things like generative AI. We could sort of start to see like what some of the applications are, like people are using chat GPT to write simple things. But I guess
47:45
my question for you is like, when you see new technologies emerge, it's kind of hard to sort of understand and foresee all the possible implications or where a particular technology like ethernet will take off. And so yeah, my question is like, how do we, as I guess young
48:04
researchers who are going to hopefully change the fields that we're in, sort of be able to see, you know, where things are going at a particular trend, like things like generative AI? Well, there's irreducible uncertainty about questions like that. I was a venture capitalist for 10
48:22
years. And there you're putting your money on decisions like that. And I learned again how hard it is to make early decisions about where to go. In my case, I fell under the thrall of Butler Lamson and Alan Kay, and they were in favor of distributed computing. So I was too. And so in
48:45
ethernet, you can see a predisposition toward distribution. We didn't add extra wires in order for the packets to take turns. We figured out a way to do it with the wires that we had, because we wanted to maximally distribute the network to be loyal adherence to Butler's
49:02
principles of distributed computing. When I gave my course at Stanford, I called it distributed computing. I didn't make that term up. I got it from him. So choosing your mentors carefully is an answer to that question. Find someone who you think you can believe in, whose math seems to have
49:24
some future, and adopt them. With that lame answer, thank you for your time. I'm going to finish 41 seconds early. Thank you.