No Way Back (5 page)

The second problem with P-Scan – and this one is non-technical – is equally serious in my mind, particularly since I’m supposed to be the ‘business person’ in the
room. No one has any idea how to make money with it. The product was born in the burst of enthusiasm that accompanied the rise of social networks like Friendster and Facebook. Everyone in the world
was putting their private photographs online, on the Internet, thereby making them public. Wouldn’t it be interesting (or so the thinking apparently went in the Tao Software boardroom,
perhaps as cannabis smoke wafted under the door) if a computer program could automatically identify everyone in a photograph? That way, you could search for photos of yourself, or for friends, or
for family – no matter whose camera took them, and no matter who put them online. A good idea, and interesting... except for that small, nagging problem that no one is willing to pay for such
a service.

These are the main points I glean from Darryl’s lengthy description of the product that he and the Tao engineers have built. After Darryl continues for some time about the beauty of
Tao’s latest algorithm, about how it translates photographic pixels into a 1024-bit hash, and not a 128-bit hash; about how Tao’s algorithm can ‘gridify’ a scan at 1/10th of
a millimetre resolution – after he tells me all this, and drones on for what seems like eternity, he finally turns to me and says: ‘So let me show you.’

Even if the software doesn’t work, at this point I feel an enormous relief, simply that Darryl has stopped talking. Randy must feel it, too, because he nods vigorously, like one of those
bobble head dolls on the dashboard of an old Dodge El Camino.

Darryl taps his keyboard. Rows of small colour photographs appear on-screen, like a high-school yearbook. I recognize most of the faces as Tao Software employees.

‘Choose one,’ Darryl says.

‘All right,’ I say, and point to the screen. ‘That one.’

I have pointed to Rosita, the heavy-set trouble-maker from the lunchroom this morning.

‘Good choice,’ Darryl says. ‘Lovely Rosita.’

He presses a key. A yellow square appears around Rosita’s head. The image is progressively enlarged until her face fills the screen. It’s blurry, not particularly well-photographed.
Blown up to this size, the image is barely recognizable as Rosita at all.

Darryl clicks another key. The image is transformed into boxy pixels, various shades of grey. As if the P-Scan software is trying to distil the most important aspects of the picture – the
essence of what makes Rosita look like Rosita – it selects some of the blocks, and highlights these in yellow. These yellow blocks mostly come from her cheek and jowl, and from a series of
blocks stretching across her shoulder, as if the P-Scan software has determined that Rosita’s weight and width are the singular aspects of what makes her who she is. I think silently to
myself that we must be very careful about how we demonstrate this product in public, particularly to women.

The remainder of the photograph fades, leaving behind only the series of yellow blocks, like a photographic signature.

The word, ‘Scanning... ’ appears on the screen, and then, below it, come quick bursts of text: ‘DMV:Alabama... DMV:Alaska... DMV:American Samoa... DMV:Arizona... ’ and so
on – through the US states and territories.

It’s probably at this point that I should mention what makes Tao’s software unique. Facial recognition is, in itself, an old art. Programmers have been doing it for years. You can
generally match any two photographs of the same person, assuming that you first tell the computer who is who. But Tao’s P-Scan does not require any sort of ‘who-is-who’ directory.
Instead, P-Scan uses the entire Internet as its directory.

So, when you ask P-Scan to identify a person in a photo, it sifts through millions of images on the Internet – from both formal sources (government driver’s licences and passport
photos), and informal sources (wedding announcements in the
New York Times
, magazine photos, or even private personal Web sites).

The idea behind P-Scan – and it is audacious, I have to give them that – is that any person, in any photograph, should be identifiable. Point to a photo, and say, ‘Who is
this?’ and let the software search the Internet to figure out the answer.

On-screen, the cursor flickers as the various databases are scanned. During the search of ‘DMV:Maryland’, the computer pauses, and dings a soft musical chime, and displays:
‘Possible Match’: along with a photograph. It’s a driver’s licence photo of a Maryland woman, with the same wide face and Hispanic complexion as Rosita. But it’s
clearly not the customer service person who works at Tao, and the computer seems to realize this, because it immediately appends: ‘Match probability: 48%’ and resumes scanning.

This process continues through the various state Department of Motor Vehicles databases. Simultaneously, I notice, P-Scan sifts through other databases, in parallel: ‘Flickr.com photos...
NBC network news... Facebook.com... Poughkeepsie Register... ’

It’s an amazing demonstration, really – hundreds of image sources, possibly millions of images – zipping through the computer’s memory, being compared to a mathematical
representation of Rosita.

I’m about to comment on this, and ask Darryl something like, ‘How many images can it process per hour? How many per day?’ – because those are the units of time I think
are applicable here – hours or days – and I am sure that it will take at least an entire day to process Rosita’s image and to find her needle in the haystack of the world
Internet.

But before the words can form on my lips, the screen goes black, and then two images appear, side by side. On the left, the grainy image from Tao’s employee photo of Rosita; and on the
right, an enlarged colour photograph from a high-school yearbook. It says: ‘Rosita Morales, St. Cloud High School, Class of 2003’. It’s a picture of Rosita – much younger
and thinner than I know her – with neatly coiffed hair, sitting primly with clasped hands, in front of a fake sky-blue background.

‘Ta da!’ Darryl says triumphantly. ‘It worked!’ He sounds quite surprised that it did.

‘Holy shit,’ I say, softly. I’m not much of a technologist – can barely use a spreadsheet, I admit – but this is one of the most amazing software demonstrations
I’ve ever seen. ‘How did it—’ I start.

Darryl launches into yet another description of his software, beaming like a proud father. ‘Brilliant, right? Well, that was pretty lucky, to be honest, because we happen to have a lot of
Florida high-school yearbooks in the database. But if you chose an employee from Oregon – you know, like David Paris? – we would have been S.O.L., because that state is insane, I mean
really insane – they won’t give us DMV access. They won’t give
anyone
DMV access. So we’re really at the mercy of the data sources. And then there’s the CPU
problem. The more data, the more CPUs you need. That’s why you gotta run this as SAAS from a data centre.’

‘But still,’ I say, not quite sure what he’s talking about, and also knowing it doesn’t really matter. This is a good demo. Lots of sizzle. Even if there’s not much
steak. Sizzle sells. Sizzle gets contracts signed.

Randy looks to me. ‘Jim, I just want to say, one more time for the record, that it’s a very early alpha release. We only get 80% accuracy. That’s it.’

I make a snap decision. ‘I don’t care. Let’s do it.’

Randy looks wary. ‘Do what?’

‘We’re going to show this. We need to get cash in the door. And the only way to make people sign cheques is to show them something they can touch.’ I indicate the computer.
‘Can we bring this thing to a meeting?’

‘It’ll run anywhere,’ Darryl says. ‘We just need an Internet connection.’

‘Jim—’ Randy starts, sounding as if he’s about to protest.

I turn to him. Something in my face tells him not to.

‘What?’ I ask.

‘Nothing.’

Darryl says, ‘I can have it ready for tomorrow.’

‘Do it,’ I say.

When I find David Paris, he’s in the kitchen, making popcorn. He’s bent over the counter, with his nose pressed against the microwave glass, staring into the oven
with the concentration of a warden counting prisoners.

‘David?’

He turns. ‘Yes?’ He looks guilty. ‘I was just making popcorn. Do you like popcorn, Jim?’

‘No,’ I say. ‘Tell me how you plan to make money.’

‘Money?’

‘With our product. You’re the marketing person here. What’s the marketing plan? How do we actually make money?’

‘Oh, Jim,’ he says, with a strained uncomfortable look, as if I am a simpleton and he doesn’t want to embarrass me, not here in public. ‘We won’t be making money
for quite some time. Quite some time.’

‘How long, about?’

‘Well... ’ His voice trails off. He shrugs. ‘It’s hard to say.’

Ding
, says the microwave. Time is up. David reaches in, takes out the popcorn, and carefully opens the bag, mindful of the scalding steam that puffs out when he prises the paper apart.
He turns to me. ‘Why do you ask?’

‘Why do I ask... how we’re going to make money? I don’t know. Just a passing fancy.’

‘Jim, you’re not from this industry, are you?’

‘What industry?’

‘Social, Jim,’ he says. ‘Social. Everything is social right now. It’s the new thing. Facebook. You use Facebook, don’t you?’

‘No.’

‘Well, there, you see,’ he says, as if he’s just proven his point. He reaches into the popcorn bag, munches on a handful, licks his fingers, reaches back into the bag, then
remembers to offer me some. ‘Want?’ he says, holding it out.

‘No,’ I say. ‘Maybe I’m not making myself clear. Who is going to pay us money for our product?’

‘Oh,’ he says, ‘I don’t think anyone. Not right away. But if you build it, they will come.’

‘Who will come?’

‘
They
,’ he says, looking around the room, as if
they
might be in the kitchen.

‘I don’t think anyone is going to come,’ I say. ‘And if they do come,
they
is going to be a bunch of kids who don’t have any money. You can’t run a
company if you don’t make money. You are aware of that fact, aren’t you, David?’

‘That’s very old-school,’ he says, smiling. ‘That’s not the way people think nowadays.’

‘It’s the way I think nowadays.’

The tone of my voice registers somewhere in the deep recesses of his elfin brain. He becomes meek and obsequious. ‘Very good, Jim. Very good. Tell me what you have in mind, and I will
implement it.’

‘Nothing,’ I say. ‘I don’t have anything in mind. Not yet. But we need to figure out a way to make money with the product that we built.’

‘Very good,’ he says, nodding. ‘Very good.’ A little kernel of popcorn is stuck at the corner of his lips. ‘I’ll start thinking about that. How to make
money.’ He points to his head, squinting and nodding. ‘How to make money... How to make money.’

He continues to mumble the phrase, and I leave him there, with his popcorn, to consider his new mantra in peace.

I stay until six.

When I call it a day, and step into the parking lot, with my suit jacket slung over my shoulder, I’m struck again by the Florida heat. Three steps to my car, and I’ve broken a sweat.
Four steps and I’m soaked. By the time I climb into the seat of the Ford, my hair is plastered to my forehead, and my skin is red and blotchy, as if I’ve spent the day working at a
smelter and not a software shop.

I crank the air, and drive home.

I call it home, but I’ve never seen it before. When I won the job at Tao, Libby and I agreed that I would take a one-week vacation without her. I flew from Palo Alto, where we live, to our
cabin in Orcas Island, just off the coast of Seattle. I spent the week fishing and thinking, in solitude. While I was there, Libby preceded me to Florida. She found us a house to rent, and prepared
it for my arrival.

It doesn’t sound very romantic, or very fair, and it’s probably not. But turnaround jobs can stretch for twelve months, without vacation or weekends. The days last fourteen hours.
The pressure is non-stop. You need to arrive at the company ready for work. It helps to have a few days of quiet under your belt before you start. Libby and I are a team, and she did her part so
that I could do mine.

So I flew in on the red-eye last night, from Seattle to Atlanta, and then to Fort Myers, and went straight to the office this morning. I haven’t seen the new house. I haven’t seen
Libby. The last time I saw my wife was seven days ago, when she dropped me off kerbside at SFO, kissed me, and told me to have a good time on my private vacation without her. I don’t think,
by the way, that she really meant it.

While I was on that vacation, Libby found us a house. Not a particularly nice house, she warned me – not really our style – but a house that would suffice for a temporary assignment.
And a house that happened to be ridiculously convenient – just ten minutes from my office at Tao.

I follow the directions on my GPS to the house. Minutes later, I turn into a deserted cul-de-sac, and pull the Ford into a gravel driveway. I see my wife right away. She is in the front garden,
on her haunches, digging with a trowel in dark earth.

When Libby hears my tyres pop the gravel, she looks up. She’s wearing a wide-brimmed linen hat, a yellow sundress, rubber clogs on bare feet. I climb out of the car, stretch my legs, slam
the door with the bottom of my shoe. I walk to her.

Every time I see my wife after an absence – even if only a day – I think to myself: How did
I
manage to get a woman like that? She is fifteen years younger than me, which
makes her thirty-something – just old enough not to be embarrassing when we show up together at a dinner party. She has chestnut hair to her shoulders, pale eyes, a pretty face, and a tall,
lanky body forged by regular gym workouts and a relentless discipline that I used to find charming but now think just a bit ruthless and scary.

I don’t know what I’m expecting Libby to do when she sees me – maybe throw down the trowel, jump up, and hug me with mud-caked gardening gloves? – or, at a minimum, smile
that awkward, toothy smile I fell in love with so many years ago? – but she does neither. Instead, this is what she does. She remains kneeling in the vegetable patch, and regards me
curiously, as if I’ve returned from a ten-minute trip to the grocery store, and not a seven-day absence.