9½ things Wolfram Alpha doesn’t know
So, after months of anticipation, Wolfram Alpha is finally here. And I don’t know about you, but I’ve found it a big disappointment.
I mean, obviously it was never going to slay Google on its first day. But after watching Stephen Wolfram’s pre-launch screencast I did believe it was at least going to be a credible alternative information source, offering authoritative and structured answers in a way no traditional search engine could aspire to.
Sadly, now Wolfram Alpha’s here it turns out that it doesn’t bloody know anything.
Depth rather than breadth
All right, let’s be fair: it actually does know quite a lot. If you try out the test queries, and your own variations on those themes, you’ll see that, actually, there is a remarkable amount of information stored up in the back-end.
But as soon as you start to stray outside of those specific areas of knowledge – and start asking it about things you’re actually interested in – the experience goes rapidly downhill.
Here are ten sample queries I’ve tried. To try to be fair, I’ve stuck to the sorts of information Wolfram Alpha is supposed to handle (as described in its FAQ): systematic, public, factual knowledge. And, just for comparison, I’ve made the same queries in Google and clicked on the first result. These were the results:
1. “House prices in Kentish Town”
Wolfram Alpha has never actually heard of Kentish Town. Maybe its existence isn’t public knowledge. Whatever the reason, this makes it pretty damn useless if you’re, say, looking for a new house within easy commuting distance of Dennis Towers.
The first hit on Google was a website detailing the asking and selling prices of houses sold in Kentish Town since 2000, along with trends and reports of how long they remained on the market.
2. “Biggest selling single of the 1990s”
Nineties chart-toppers may not be to everyone’s taste, but the facts and figures are (if you’ll pardon the expression) a matter of record. Sadly, Wolfram Alpha has none of them. It did, however, tell me that Shakira’s surname is Ripolli. Wouldn’t Graham-Smith suit her better?
The first hit on Google was a Wikipedia page telling me that the single in question was “Candle in the Wind 1997”, which reportedly sold a sickening 37 million copies.
3. “Newspaper circulations”
When I entered this as my query, Wolfram Alpha’s first guess was that I might want to know about Sports Illustrated. Second guess: The Idaho Statesman.
The first hit on Google was the Audit Bureau of Circulation. From here I had to engage in a little navigation, but after six mouse clicks I was rewarded with a list of UK national papers and their average net circulation.
4. “Bicycle theft in London”
As a recent victim of this heinous crime, I’ve been wondering how many bikes are stolen every day, where the hotspots are and where, if anywhere, they resurface. Unfortunately, to get any response from Wolfram Alpha at all I had to generalise this question down to just “bicycle”, at which point it told me that the modern bicycle was invented by John Kemp Starley.
The first hit on Google was a blog post advising me that expensive bikes are less likely to be stolen than cheaper ones. Not much consolation there.
5. “Eurovision winners”
After Norway attained the highest Eurovision score ever on Saturday, I started wondering about previous winners. Which countries had done best, and which worst, over the years? Wolfram Alpha guessed that perhaps I wanted to know about either the euro or the AAA Vision aircraft.
Google once again led me to the Wikipedia page on this topic, which offered all the information I could possibly want (and far more beside).
6. “CPU die sizes”
The original Pentium was built with a 0.8µm process. Now we’re down to 45nm chips, with 32nm on the horizon. Can Wolfram Alpha show me a graph of how die sizes have shrunk over time? No. It thinks “CPU” is an airfield in California.
Google’s first hit was an online article defining die size. I didn’t get the graph I wanted, but at least we were in the right industry.
7. “Web browser market share”
Admittedly the data on this one aren’t as solid as some of my other searches, but I was optimistic Wolfram Alpha would know something about web browsers. It doesn’t, but it did invite me to leave my email address so it could alert me when it does learn something on the topic.
For the third time, Google led me to an extensive and informative Wikipedia page, complete with a pie chart, a stacked line graph and numerous tables detailing the specific figures from the various data sources.
8. “UK unemployment 1980-2008”
For once, Wolfram Alpha did understand what I was asking… but alas it was unable to help due to “insufficient data available”. A few follow-up searches revealed that this is because it doesn’t have any data for any of those years. Oddly, though, if you leave off the year it gives you an estimate for 2008 (5.5%, if you’re interested).
The years threw Google off. It took me to an Amazon listing for a live DVD by a band named “The Inmates”, recorded in 1980 and released in 2008. A rare fumble.
9. “World Cup runner-up”
I tried several variations of these search terms, but Wolfram Alpha could tell me nothing more useful than the volume of a standard cup. Apparently, in the US it’s 0.2366 litres – I had to click onward to find the size of an imperial or metric cup. So much for tailoring results to your location.
Yet again, Google went straight to Wikipedia, this time to the FIFA World Cup page. I had to scroll down the page quite a lot to find the information I wanted, but it was there. (It was France.)
10. “Chemical symbol for tungsten”
“Wolfram|Alpha isn’t sure what to do with your input.” Oh, come on – really? Oddly, when I tried knocking off the word “chemical” it did yield the correct result, so I’ll give it half a point for that one.
With Google I didn’t even have to click on a link: the answer was right there above the search results, along with the source.
Do you have to ask? Wolfram Alpha is clearly a very clever bit of programming, but if you actually want to learn something it’s next to useless.
All the same, we shouldn’t underestimate its potential. The engine that combines and presents arbitrary information in a comprehensible way is undeniably powerful. If it can only extend its ambit beyond institutional statistics and scientific formulae – if, in other words, it can start answering everyday questions about everyday topics – it could yet leave Google looking laughably primitive.
But before that can happen it needs more data. Lots more data.