Original URL: http://www.theregister.co.uk/2006/08/25/colly_myers_interview/

Are Google's glory days behind it?

Colly Myers on the rise of AQA

By Andrew Orlowski

Posted in Applications, 25th August 2006 14:42 GMT

Interview Answers service AQA is two years old this summer, and finds itself in the happy position of not only being profitable, but something of a social phenomenon in its home country.

A book based on the service, The End Of The Question Mark is due to be published in October, drawing the questions Britons ask, and the answers AQA gives them. Not bad for a company that still has only nine full time employees.

What AQA allows you to do is text in a question and receive an answer for a quid. This might strike US readers as expensive: it's nearly two dollars (or four days of the San Francisco Chronicle) for a few lines of text at today's exchange rate. But Britons love texting, and arguing, and AQA's combination of canny marketing and the quirky charm of AQA's answers have proved to be a hit.

But where AQA particularly interests us is how its success poses a challenge to a lot of the Californian-inspired orthodoxy about search engines, and Silicon Valley's latest hype of fetishising "amateur" content.

These are strange times indeed when an AOL web executive must defend his decision to pay former volunteers real money for their labours. Actually pay them - so they can help feed their families? The horror of it!

Founder Colly Myers had plenty to say on this, in typically no-nonsense style, when we caught up with him recently.

AQA served its 3 millionth answer recently, notching up the last million in four months. The previous million took seven months, and the first million took 19 months, which gives some indication of its growth ramp.

AQA's owner IssueBits has been profitable since last October, says Myers, and he thinks the market is young and there's plenty of opportunity to grow. AQA doesn't have the field to itself - 82ask also caters to the curious texter - but it is in pole position.

Myers seems particularly proud of the infrastructure: AQA uses around 500 researchers to answer double the volume of queries it did before (the actual composition of the research staff varies, as they drop in and out of work).

The internet's "search business" has had saturation coverage in the last couple of years, typically in hyperbolic terms. The subtitle of one recent book suggests that the company "rewrote the rules of business and transformed our culture" - that's a book that Google liked so much, it bought hundreds of copies for its staff. But very little of the coverage has highlighted the philosophical and practical flaws of the web search business.

AQA's researchers don't use Google (or Wikipedia, which we'll come to). And internet search is a business that may already have seen its best days, Myers suggests.

There are several reasons to support this view, and in some cases they're interrelated. One is that Google is finding it increasingly difficult to maintain its index. It's ignored the second law of thermodynamics.

"It's a well known aspect of man and machine systems. Complex systems with no control fall over. Every example of it you can think of falls apart. With databases, data that isn't pruned becomes overgrown. Entropy sets in when complexity gets out of control.

"A lot of the search engines' index is junk, and although they have a lot of clever people, they can't prune it manually. And they have a lot of powerful technology too, but they just can't stop it.

"We're looking at the prospect of the end of the growth of search. Like Microsoft, search won't be going away anytime soon, but the search engines' power will weaken for several reasons."

The difference with AQA, he suggests, is because it's paid for its answers, and not for advertising. So keeping up the quality of the inhouse database is vitally important as it affects the quality of the service.

"So much so," he says, "that it is worth spending money on improving the data in the database. There is not the same incentive for search engines."

"At AQA we don't use Google. We don't use Wikipedia. We have good researchers who know where to go - to the primary sources."

Nor does Myers see gimmicks solving Google's massive accretion of what's called "Goobage".

"People come up with ideas like tags, metadata, but am I going to take the trouble to do that on every document I create? Are ordinary people? Google should pay people for adding the metadata.

"That's very valuable information to Google. Why should people donate their time for free to help Google out of a hole?", he asks.

Because they think they're helping God? It's a fair point. When simple software can create 100 new weblogs for your junk content in 24 minutes, entropy is going to out pace the most devout, Google-worshipping tagger (those spam blogs use tags too, of course).

There are more reasons that the web is in a big heap of trouble when it comes to answers. Once you've got the MySpace habit, you rarely leave the site. Myers explains:

"The 18 to 24 age group today is not using search so much. Why? People know where to go. They have MySpace - they know where the community is. They know where YouTube is. People need search engines less and less, because they don't need to use search as a portal."

This hunch seems to be confirmed out by two pieces of evidence.

MySpace is the number one destination in the US, but MySpace mail, the company's European VP said recently, is number four. So people search from where they already are, within these vertically integrated sites, and there's less value to having Google as your home page - or leaving MySpace.

(Perhaps sensing which way the wind is blowing, Google recently paid almost a billion dollars for an advertising deal with MySpace).

Another clue is this traffic report from HitWise, which saw Google (in green) gain a lot of traffic the day MySpace (in red) crashed. Click the graphic for more details or here for more details.

How Google gained from MySpace's downtime That's quite a spike.

But aren't teenagers fickle, we wondered? So they can't bet on commanding an audience?

"If they don't deliver value, no - of course not," he agrees. "The community provides the validation and without the community there's nothing. Consumers have an tremendous amount of power now and can switch very easily, so unless you add value they'll go somewhere else."

"But we could start to see browsers with the bookmarks for MySpace and YouTube preloaded, or some other form of bookmarks. That's very powerful."

There's a third powerful reason why web search isn't as good as it was.

"The 80:20 rule also applies here," he says. "Twenty per cent of the documents provide most of the value that people want. The other 80 per cent add less value. Now what the search engines are doing is indexing stuff that increasingly adds less value. Meanwhile, more and more data is disappearing into the 'dark web'. And by nature a dynamic web service is not 'indexable'."

The thought struck us, round about here, that the idea of the internet as a great leveller, bringing information to the masses, might just be about 100 per cent wrong. Or at least highly overstated.

As Google's fate is increasingly linked to amateur or highly dubious content like Wikipedia - and the many sites who mirror or scrape its content - the contrast between high quality research and low quality web search seems ever more apparent. The popular fetishisation of amateur web material is a peculiar belief as it needs to suppose that paid information doesn't exist, and isn't better. Nonetheless, aren't we seeing the emergence of two worlds of information - one low grade, amateur, and beset by entropy: the open web - and the other of high quality? Of course those of us with membership cards are laughing all the way to our libraries' expensive database collections - and we can afford to pay for quality.

Spare a thought for the rest of the world, though. Access to the web begins to look less like a blessing, and more like a curse.

Which brings us to the vexatious subject of Wikipedia.

"For me Wikipedia is a classic example where you appear to solve one problem, but don't really do so, because you create another problem. In the case of Wikipedia it created a community to generate information but failed to find a way to authenticate that information. So close, but still so far.

"I am doubtful that community based systems will be the basis of sustainable solutions. I suspect that a commercial imperative is necessary. It's the same with all these volunteer systems, they're idealistic, but no one comes up with a way of paying people."

Now we were extremely skeptical that AQA didn't drink from the Wikipedia well. But in a couple of tests, pitting it against rival 82Ask, AQA drew from more original source material, while 82Ask hit the Wikipedia. It was still web-based, but made for a better answer. AQA was much faster, too.

AQA's success arrives at a time when there's a lot of utopian talk about how new production "models" may be being enabled by technology networks - often using home-based workers offering their services for free. AQA also taps into a global, networked workers - but pays them.

Much of this is misplaced, and quite fanciful, the result of academics who've arrived at a "model" simply casting around for evidence that looks like it might fit their favourite shape, such as Yochai Benkler. When academics fasten onto a favourite shape, no amount of reason can persuade them otherwise.

But there's also a darker side. Amidst a lot of hooey about "collective intelligence", there's the thought that we're enabling virtual sweatshops. Amazon.com's Mechanical Turk is one example of this kind of business - bringing together employers who have piece work, often mundane and repetitive, with people prepared to do it. It's been called a virtual sweatshop.

Colly Myers rejects the idea that AQA may be depressing wages.

"What we do at AQA is bringing people into the workforce - and in many cases they wouldn't be otherwise working. Students can work, but two thirds of our researchers can't - they're at home. We can do this because we have a business perspective, we have revenue.

"So the technology allows us to do so very efficiently. We're much more efficient than a call centre or a support centre. And they're a lot happier because they work from home and can choose when they work - there's unimaginable freedom," he says.

He agrees there is at least one similarity between Mechanical Turk and AQA in that they both use humans to do tasks that are too difficult to be done by computers, both have a task that can be dispatched and completed over a network, both pay for the service, and neither schedules any resources.

But Mechanical Turk is a generic clearing house, which leads to its major flaw - lack of quality control.

"The very nature of this genericity leads to its key weakness in that it does not have adequate mechanisms to qualify its human resources to any specific task. It does try to define various types of qualifications but they are necessarily weak and fail the test as a true qualification for most non-trivia tasks," he says.

"As a consqeuence, the quailty of their work against any task cannot be very high, and consequently a user of the Mechanial Turk system cannot afford to pay a great deal for the service.

In other words, Amazon can't do the tests that AQA performs to qualify researchers, or the reviews the senior researchers do of their answers. And there's no management structure to monitor and train people.

"Continuous improvement is vital since without continuous improvement any business is a dead duck over time," says Myers. "I think that it is something close to a universal truth to say 'that you get what you pay for' in the cost vs quality debate."

He rejects the idea that Mechanical Turk participants are in a "sweatshop", because they can leave at any time.

Finally, the former Psion MD and Symbian first CEO remains as committed to data integrity as he always has been - he keeps paper copies of his contacts book under his bed he says, although he may have been being metaphorical.

We mentioned David Rosenthal's description of engineers who share these values as essentially pessimistic in their approach. They need to factor in whether the bridge will fall down.

"Someone asked me what I was at a family gathering, recently. I said I'm a pessimist. They replied, no you're not, you're a cautious optimist."

Psion, the company that invented the PDA, had a thing about never losing your data. That's more than you can say for us. After an hour of recording our conversation for posterity on a Nokia, the phone simply looped round and deleted the recording. It wasn't a Symbian Nokia. ®