Is Google the end of history?

Today's Guardian has an interview with Esther Dyson, editor-at-large of CNet. Here's a taster:

Technology Guardian: With the rise of Web 2.0, we're seeing huge growth in the number of startup companies and funding available. But what do you do when a company like Google squashes your business before it's started?

Esther Dyson: Really, the big question is, is Google the end of history? And the answer is that it's not, because the problem is that Google doesn't understand meaning: all it does is index text on the web - it knows nothing. For example, Time Warner acquiring AOL means something very different from AOL acquiring Time Warner - but Google couldn't understand the difference.

So the next big technology will be pattern recognition, which helps us understand what's going on and allows us to understand meaning. Right now a search engine is just a database; it isn't able to connect the dots. That could change if we had better natural language understanding, and Reuters, for example, is already pretty good at some of this. As we move further into TV and photography online, it's going to get more important. I'm an investor in a company, Powerset, which is working in this area.

Comments

[A. PERLA]

"Really, the big question is, is Google the end of history?"

Really, the question is, "Can Esther Dyson get media attention without making such stupid remarks?"

Oh, thank God, she thinks it ISN'T the end of history! Wow, had me worried, there.

Anybody with the least bit of experience in ITC knows full well that product / technology obsolescence is waiting likely just around the corner, in some garage, to be developed and commercialized. Google has managed to obtain a quasi-monopoly in net search engine technology that is it is milking to the hilt with outrageous publicity pricing. (Auction publicity spots to the highest bidder! You gotta be kidding! But, no, they're not. We are told they are "clever" people. Yeah, right, they said the same thing about Bill Gates practiced monopoly pricing to sell MS-DOS.)

For once, I hope Microsoft’s new internet search engine comes out soon and sweeps the floor. Comeuppance can be soooo sweet!

[A. PERLA]

Dyson: "So the next big technology will be pattern recognition ... "

Perhaps. But, given the problems of the technology, I wouldn't go rushing into this field investing willy-nilly.

Pattern recognition in a critical mission application is NOT doing all that well. The technology employed to recognize facial characteristics of people passing a barrier (in an airport) is having its teething problems. Nothing guaranties that the technology can do the job well enough to be counted on.

Natural Languages is a better bet. The Boolean language of today’s search engine technology is not bad. However, it is not commercially an acceptable way for Joe Bloggs to search the net because it would require the understanding of some fairly simple Boolean concepts. Apparently, even that is too much to ask of most people with a high school education.

So, the experts are saying that "Natural Language" is the way to go. Meaning simply that the technology must capture the babble that people write as a contextual search and ... make sense of it.

This is, I would suggest, a flagrant example of the dumbing down of education. Browsers on the Web can’t seem to write properly a phrase specifying their search criteria, so the technology must interpret what they mean. (Example of a Web Natural Language search, "Hey, dude, gimme all ya got on sellin machine guns and leave out all the stuff about gettin import permits.")

This is "natural"? Will wonders never cease. ; ^ )

[AC Mitchell]

I'd be interested in what she actually means by 'pattern recognition' - how she expects that to manifest itself as a technology. Is it patterns of what we look at / for ("You like gin; you like PG Wodehouse; computer posits innate connection between gin consumption and novels of the 1920s")? How would this information be used other than the rather obvious, advertising?

Her point that Google 'is just a database' seems even more confused, firstly because it isn't and secondly because the one thing databases do model is the relationship between things.

[A. PERLA]

Mitchell: "Her point that Google 'is just a database' seems even more confused, firstly because it isn't "

Well, as long as it is a key-word look-up and reference to the original document, then it is a "data-base". Or, enormous index, if you like.

It certainly must employ data-base software to manage site indexing.

And, frankly, there may be a legal issue involved. Google, I suspect, may have to request permission to index information that is proprietary. There is an issue, perhaps, of intellectual property rights. But, neither do I really know since I am not a IPR expert.

[AC Mitchell]

Index is probably the better word - the point I was trying to make there really that databases already show us relationships between things, whereas a search engine gives us information in a much 'flatter' form. Ms Dyson seems to conflate 'database' and 'index' as both being things which do not show us meaning, which is inaccurate as relationships between things are shown very well by databases. This is probably so technical as to be off-topic, but to explain a bit more -

Let's take her example of the Time Warner takeover of AOL. Around that time a Google search for either "AOL" or "Time Warner" would probably have rendered up a fairly similar set of top hits, perhaps with the first being the corporate website of whichever you searched for and the next, let's say, three being the same news articles (BBC, CNN and Sky) about the imminent takeover. Meaning is not, in this case, evident from the list of articles, because whether you search for the company taking over or the company being taken over you get more or less the same results.

Now let's imagine I keep a database of all the listed companies in the US since 1997. There are two tables; one holds data on companies currently listed, the other data of companies taken over. There is, in this model, a clear heirarchy; each company acquired will have the name of a currently listed company in a field attached to their data. A currently listed company can have many acquisitions, but an acquired company will only have one company in their 'owned' field.

So, whilst my example is obviously lots less useful than the Google search, it does 'join the dots' in the way the flat index doesn't. Re-reading the quotation it sounds like Ms Dyson is talking more metaphysically than anything else (or perhaps she herself is not sure what she means)

Re: IPR, I believe people have to actively consent to have their pages spidered by Google. So it's not so much a case of Google requesting permission from people as of people requesting Google to index their site.

[A. PERLA]

Mtichel : "Re: IPR, I believe people have to actively consent to have their pages spidered by Google. So it's not so much a case of Google requesting permission from people as of people requesting Google to index their site."

When one sets up a site, "referencing" is the final bit by which the site is submitted to a number of engines for indexing. So, yes, you may very well be right, a site requests to be indexed.

But, the "bots" go out dragging the Internet and they do not necessarily ask permission to index a site.

Look, I'm no expert, but I sense the issue is there. Perhaps, as well, it is a tempest in a teapot. I wouldn't know.

I am troubled enough that Echelon is copying my voice / text communications. Now, THAT really scares the hell out of me. And, get this: When Airbus was about to capture a large order in India, the people at Echelon warned someone in the administration (not this one, but a past administration) who promptly called Boeing to tell them they had better move quickly. Which Boeing did to capture the contract.

[Chuck the Lucky]

If the next big thing in search engines is "natural language" then I certainly hope that the next next big thing is some sort of reasoning ability. A search engine that can give you millions of sites that deal with the exact subject you are interested in will not be much help if it can't figure out that you are more interested in an intelligent article on say, globalization than in one that blames the evils of globalization on the reptilian aliens from the planet Moobar (hmm... sounds like something the New York Times or The Gardian might put out if they can tie Bush and the Jooz into the story).

[A. PERLA]

"If the next big thing in search engines is "natural language" then I certainly hope that the next next big thing is some sort of reasoning ability."

Great idea. I'm tired of reasoning too ! ;^)

[Web Design Kent]

Correlation means very little on very large data sets causation or otherwise. On really big data sets the probability that data will correlate while having no link whatsoever increases. With a large enough data set the probability of random bits of data correlating is almost a certainty.

[Web Design Kent]

It is really difficult to say how this will pan out. The Web 3.0 will be an exciting semantic thing.

[mcx]

Really google has dominated other search engines. It's simple. It's clean. It's fast. It gets you the information you want. Google has a problem that is creeping up on it. It certainly must employ data-base software to manage site indexing.And, frankly, there may be a legal issue involved. Google, I suspect, may have to request permission to index information that is proprietary.
Sureshot Intraday Tips || Sureshot Option Tips