Calculated risks: Will algorithms make business boring?

Ever found yourself wondering whether your boss is human?

Haven't we all.

Now, with the march of artificial intelligence and robotics, it's becoming an increasingly valid question.

Algorithms - problem solving computer programmes - are, to put it bluntly, getting much better at doing our jobs than we are.

And it's not just in the tech sector where computers are becoming king.

Last year the LA Times published an article about an earthquake which was written by a computer programme.

School children in South Korea are being taught English by a machine called Robosem, and even the so-called oldest profession faces competition - a US firm called TrueCompanion claims to have developed the world's first sex robots, with the most basic model costing just $995 (£645).

But is there a danger that this brave new world is going to be a bit, well, dull?

Movies by numbers

The entertainment industry has already adopted an algorithmic approach to working out what we want to watch.

When TV and movie streaming service Netflix decided to start commissioning its own material, it turned not to Hollywood veterans, film critics or media forecasters, but to algorithms and user data.

A data trawl of the most-watched and loved content streamed by Netflix customers revealed three key ingredients - actor Kevin Spacey, director David Fincher and political dramas produced by the BBC .

So the firm commissioned a remake of 1990 BBC political thriller House of Cards, starring Kevin Spacey and directed by David Fincher. It became the first ever web series to win a prestigious Emmy award, with the first series receiving nine nominations.

"House of Cards is the most famous example of a data driven approach to creativity," said Luke Dormehl, film maker and author of The Formula: How Algorithms Solve All Our Problems, and Create More.

"Netflix commissioned an unprecedented number of these episodes - and paid upfront - rather than going down the traditional route, where you shoot a pilot, present to executives etc."

It is perhaps no wonder the company keeps its lucrative viewing figures and ratings information such a closely guarded secret - it has never made them public.

London-based firm Epagogix offers an AI analysis of movie scripts to predict how successful they are likely to be at the box office. It even claims to be able to identify "improvements" to boost a film's commercial value.

In theory this eliminates every film maker's worst nightmare - the expensive flop.

"If a film needs to make £10m and it makes £5m [at the box office] we say it's bombed," said Mr Dormehl.

"But if we had that data in advance and said, if you spend half the money making it, it will still make £5m... it opens the way for making mid-range films."

But where's the excitement, the joy of the surprise hit or chance discovery of the new?

Luke Dormehl says it opens up a world of different opportunities.

"As a film maker I think there are plenty of exciting possibilities," he said.

"It's exciting to have a script where you can tap into that data. Which scenes remind people of good times, which music cues do they like, when do they get up to make the tea....

"I'm excited about the possibility of films that change direction depending on who is watching them, or change narrative if you're not paying attention."

High tech traders

The finance sector is another enjoying the comparative stability of computer control.

"The trading floor was a very exciting place. Now it's more like a software company than a financial organisation," said Dr Juan Pablo Pardo-Guerra, assistant professor at the London School of Economics.

High Frequency financial trader Virtu, an electronic trader with computerised strategies, has only ever recorded one day of losses in nearly six years.

On its website the firm claims that its 148 employees are the "secret sauce" of its success but its own proprietary technology is at the heart of all of its trading activity.

"If we were handling the same volumes of stocks with human traders it would probably be a much more volatile existence," said Dr Pardo-Guerra.

It's nothing new - attempts have been made to automate the trading aspects of the financial markets since the 1980s - but is the Virtu success story a nail in the coffin for traditional finance sector jobs?

"The first step has gone already - the floor traders. In Chicago the trading floor was the centre of everything until a few years back when everything moved to automated," Dr Pardo-Guerra said.

"The next step will probably be analysts and people who compile data and information on companies and produce stock valuations - that will be affected by algorithms and automations."

Financial thrill-seekers will still enjoy working in the sector, predicts Dr Pardo-Guerra - but there will definitely be a change of scene.

"It's probably going to have some impact on how people relate with the market, how they find excitement in trading," he said.

"But the market is so big and complex there will always be space for excitement."

And the good news is there is still room for the human touch - at least for now.

"Even within the most analytical part of finance of markets meetings and relationships are still quite important and that's something algorithms can't do," he added.

"Some aspects can be automated - calculated or quantitative decision making. But others rely more on personal cues and networks."

Algorithms have another potentially fatal flaw - they are unable to recognise when someone is taking the tricks of the trade too far.

"People know how to manipulate prices to get more customers or make their strategies more profitable," said Dr Pardo-Guerra.

"I think algorithms present challenges in terms of identifying these processes."

Their inability to interpret sneaky human behaviour has proved a stumbling block in the surveillance sector too.

OK Computer

Dr Daniel Neyland, professor at Goldsmiths University, London, was involved in an experiment at an airport last year where an algorithmic system was installed to identify suspicious abandoned luggage from CCTV footage.

"On average there was one alert per hour by having staff looking at monitors," said Dr Neyland.

"In six hours the system detected 2654 [alerts]. It was so far off the scale of expectations, there weren't enough people to respond or even go through the data afterwards."

Among the false alarms identified by the system were shadows on the floor, cleaners' trolleys and people standing still, Dr Neyland explained.

"People are generally speaking good at picking out what's a bag and what's not - but also whether it looks abandoned," he said.

"Algorithmically processing this kind of data is challenging. What's a bag and what's not? How long has it been left for? How far is a person from their bag?"

For similar reasons one European rail network has completely uninstalled an automated security system, Dr Neyland added. It was supposed to monitor un-staffed areas for signs of vandalism and break-ins.

"The operatives looked like they had a bad response rate but the system was not picking out things they could respond to," he said.

"This is a tricky area for video analytics. You don't want false negatives where you miss the thing you are supposed to look at, but too many false positives undermines ability of the system to operate."

But the day when algorithms become the boss is closer for some workplaces than others, he warns.

"Workplaces that are already tech-saturated will lend themselves more to algorithmic governance," he said.

"It already happens in call centres. Algorithms read how long calls are, how successful they are, measurements of response time etc. That data is used in managing the work place.

"There is some suggestion that staff welcome this accurate measurement of their day - it gives them an objective element to appraisals."

Not everybody however is a fan.

"As soon as you put into place a measure it quickly becomes a target," Dr Neyland added.

"There are also some suggestions that people hate it, and that it puts extra pressure on their jobs.

"If someone is going to assess quality is it OK to reduce it to number crunching?"

Automated, unflinching and calculating with no sense of humour..... maybe your boss isn't so bad after all.

/By BBC/