Wednesday, July 17, 2013

Moving on from Clippy

"Bill Gates says software assistants can save the world," says MIT Technology Review.

Readily accessible floods of data, combined with powerful cloud computing and advances in software, should allow personal assistants to “understand” what humans are doing and actively collaborate or help them out, said Gates. Smart assistant software could even be crucial in addressing some of the world’s most challenging problems, he said, as identified by his philanthropic organization the Gates Foundation.

Gates said, for example, that software able to intelligently interpret the available data about a person and respond could improve education.

The software agents could help make decisions, too.

Gates imagined an assistant helping a Kenyan farmer using the M-Pesa mobile banking system that handles roughly 30 percent of the country’s GDP (see “Shopping via Text Message”). “When I sell my crop after the harvest, it advises me to save some of my money and even makes that deposit,” said Gates, who praised the data M-Pesa has made available for such projects

Clippy, the notorious Microsoft paperclip "helper", had perhaps just been premature, he said.

Clearly, there's something to this argument. But for now, I find the services available surprisingly primitive. Nobody has yet developed an effective assistant to help prioritize an email inbox, for example (at least commercially). Junk filters and prioritizing some senders is about all email programs do. Compare this to the confident hopes about artificial intelligence and expert systems in the 1980s. Far from super intelligent computers taking over the world, they can barely sort important work time-sensitive email from routine information emails.

And for all the discussion of predicting wants and needs, I still find one of the most common examples, Netflix suggestions, amazingly primitive sometimes. It has not yet figured out there is a man and a woman in our house and they often like different movies.

The difficulty has always been computers have difficulty with context and meaning. Brute force and massive amounts of data, such as Google Machine translation, is starting to change that. However , it looks like we're just going to see specific small-scale helpers for the time being.

That could still have an impact in education or office productivity. But expectations have also fallen.



Tuesday, July 16, 2013

"For the Last Time, Robots Do NOT Cause Unemployment"

Here's a pushback at RealClearMarkets against recent worries about technology and the job market.

Parts of the nation's commentariat have been seized, in recent months, with a nasty bout of technophobia. Technophobia is a psychological condition, but infectious. Hardly a week goes by without a new outbreak documented in another blog post or business column. To judge from the symptomatic hand-wringing the epidemic is spreading, we are on the verge of mass unemployment as work becomes increasingly automated.

As I've often said, it is clearly true that there has always been an increase in aggregate demand for labor in the past. But to argue that will continue indefinitely carries an assumption the nature of needs and wants is static. Economics is not that good at understanding changing needs and preferences. Instead, they are treated as exogenous. And that is a foolish assumption.

In fact, it's not at root an economic problem. It's an ethical and instiutuonal problem. There needs to be deeper change and adaptation in economic institutions. That more attention to issues of the "good life " given basic economic survival is no longer the prime problem facing humanity. The labor market achieved huge prominence in the industrial revolution period, dominating life in a way which would have been inscrutable to medieval peasants on the land, or aristocratic Romans. It may be less prominent in the future.