Past, Present, Future Vision of AI – Google and AAAI 2011

“And it was fun looking at the comments, because you’d see things like ‘well, I’m throwing in this naive Bayes now, but I’m gonna come back and fix it it up and come up with something better later.’  And the comment would be from 2006. [laughter] And I think what that says is, when you have enough data, sometimes, you don’t have to be too clever about coming up with the best algorithm.”

(Peter’s speaking about search algorithms, but what if you applied this to running a startup? Or life in general?)

Which would you rather have:

• An app that, based on preferences you enter about cuisine, price, experience, etc., comes up with a list of restaurants in your neighborhood that fit your criteria?
• A friend who’s been to every single restaurant in town that you can always count on to come up with a great suggestion for food, no matter what you’re in the mood for?

With some exceptions (and there are always exceptions), I’d rather go to the person first, then the app. Yelp has plenty of suggestions for coffee and wi-fi in the city, but I still check Cafe Talk. Amazon has great reviews, but I’ll often still look at the page of that one blogger who writes a 20 page article just to talk about the pros and cons of one set of headphones.

Why pick one person’s opinion over the wisdom of crowds and a 200 million dollar company?*

Simple: The person cares, and the person is a person.

*Note: Yelp is actually an exception, to the extent that its full of people. But the people all care about different things.

Products are utilities, services are solutions

First off: this doesn’t apply everywhere.

I don’t need a personal assistant to read me my emails or backup my files. I’m OK with large warehouses of computers doing that, because I know exactly what needs to be done and there isn’t anything surprising or complicated about it. When I know what I need, I look for a utility.

What separates utilities from solutions? Uncertainty.

If I’m looking for a website designer, who would think it was acceptable to hand me a copy of Photoshop, a book on HTML, and say, “here’s your app, go for it?”

A utility is designed to take things we know we need and get them. The best utilities tend to be well targeted and have a specific function. No one has to ask Instagram what he does at a party; it’s pretty ______ obvious.

But even there – how did you get introduced to Instagram?

Did a friend tell you about it? Did you read about it in a blog? Did you see someone using it?

Instagram started out as something unknown, and when it comes to unknowns, we need a personal introduction.

Even if it’s a product, start out with a service

I know the eventual goal of a product-driven startup is to build something that “scales” out–the definition of scale being that we can go out and have a picnic while our software makes monkey.

Nonetheless, I see lots of people who start out with the picnic. Oh, they may be working hard on their product, but there’s no relationship to the customer. They build the system and throw it over the wall and hope it goes viral. Starting a connection to your customers through an impersonal, on-stage launch is how you get casual interest. At the beginning, there will be no one but you to give a personal introduction to your gadget.

It can be as simple as telling the story of what it does and letting them use it, but you have to provide that service the first few times around. Even in the greatest product driven businesses, it’s people who are telling the story, not the product. The eventual goal of scale is not to stop telling the story, but to have others tell it for you.

The problem: Why is this d_mn thing taking so long?

Ever since I started working on my own, I’ve spend far more time figuring out how to motivate myself. Because I’m working in a startup, I make sure to do just the minimum amount of work to make a viable product, so that I can save time and work on other things.

And yet, I find myself staring into space, going on Google Reader, Hacker News, playing games, or anything I can find aside from work.

As I’m talking with a friend online, we make some observations:

• While the eventual result is interesting, the work I’m doing is boring.
• The reason I’m bored is because the work is easy.
• Therefore, if I want to be more productive, I need to make my work harder.

I throw out the disorderly, simple, get it done code that everyone says you should write for a prototype. I make things organized. I put all my variables into neat, sensible structures that follow design principles. I use a regular expression to replace 6 lines that I could’ve done by hand in five minutes.

All of a sudden, I’m getting work done again.

What gives?

–

I’ve talked with scores of startups over the years, and they consistently err on the side of overbuilding rather than underbuilding. Fancy prototypes take months to code, months that are wasted if your idea of a customer is the two people you told at the startup meetup you went to last week that said it looked kind of cool; people create entire frameworks to build sites that could be done in the time it takes to write a shell script (that installs WordPress).

Thus, the rule to build fast and build cheap was born.

Am I some kind of unlikely exception? Perhaps. But after giving it more thought, I realized that the need to make some things hard is not surprising, and not unique.

–

Are you working at a startup?

Why?

The pay is (sometimes) lower and the responsibilities are (often) greater. Most of the programmers I know working on their own projects are exceptionally good at programming. The hackers aren’t hanging out at the back of the bell curve. They can pass FizzBuzz, they know it, and that’s exactly why they’re finding a place where they live or die by their skills.

So it’s not entirely surprising that, if you pick people who are fairly competent at coding, so competent that they want to go to a place where they may be solely responsible for the technical success of the product they build, that they might just prefer not to always have simple, somnambulist-friendly code assignments.

Now, as with any essay that argues a point, you can poke holes in this thesis. Perhaps, you might say, the original rule isn’t against giving yourself a technical challenge from time to time, but rather as a safety railing to stop the people who want to make Pandora Craigslist Farmville with an extra layer on top so you can tag your singing cows with pictures from your last missed connection. And that’s a fair point.

What we do thrives on exceptions. I wouldn’t be able to think of a good example, but by chance, I read of one in the NYT yesterday:

When asked what market research went into the iPad, Mr. Jobs replied: “None. It’s not the consumers’ job to know what they want.”

Do what works.

So you’ve decided to strike out on your own, by freelancing or start a company. The first day you get up, you realize you’re free. No office, no working hours, no one staring over your shoulder, no constant pressure to have your butt in a chair doing nothing. Great, right?

Well, kind of. The first month works great: you’re super productive and everything gets done in two hours, and then you go  home and watch TV the rest of the day. Then your subconscious realizes that if you stay home, you can do nothing but watch TV and surf the web all day, right away.

So you go to coffee shops and libraries. For free, or the price of a latte, you can get wireless, and work with fewer distractions. [1]

There’s only one other big problem: the alarm clock.

[1] If that’s too much angling for seats for you, you can also find a good coworking space for a few hundred bucks a month.

The alarm clock

Credit

The alarm clock is, biologically speaking, the enemy. Studies almost universally seem to show that having less sleep is the equivalent of taking the stupid knob on your brain and turning it up, all day. Any machine that interrupts your natural sleep cycles can’t be good; in nature, it’s essentially the equivalent of having a (mostly) friendly bear wander by your hut at 6 am every morning; you wake up groggy, stressed out, forced into action. An entire company was started and funded around avoiding it.

So, naturally, the first thing you do, once you’re free, is turn it off. The only problem: You’re a night owl. So now you have random hours. Every day you wake up, and you’re not sure what time it’ll be.
Do you just head to the same cafe, whether it’s open for 12 hours or 4?

Cognitive Load

One of the hidden advantages of regular office hours in a regular office is reduced cognitive load. If you have to work every day, but you’re not sure where or when, you spend time thinking and processing: “Is this place open? Is it worth it to commute 30 minutes if they’re going to close at 5? Maybe I should just stay home today. I wonder what’s on my email.”

Continued for too long, this thinking drains your reserve of mental energy, before you’ve gotten to work.

With a rigid, defined procedure, none of this thinking takes place: You roll out of bed, put on your clothes, do whatever and in 30 minutes you’re in the office. And you haven’t had to make any decisions yet. Your mental energy is saved for when you get in and tackle the first work-related problem.

That’s the benefit of reduced cognitive load.

Morning Routine 2.0

Does this mean that you, the freewheeling self-employed freelancer, need to don a suit every single day, set your alarm clock for 4 in the morning, and go for a run of 1.8 miles before coming in for a coffee and commute to the office?

Fortunately, it doesn’t.

The most important thing that you need to have in the morning is not super rigid rules, but a set of algorithms that reduce cognitive load.

Most likely, in the morning, you’re making the same set of decisions, over and over again:

• What should I wear?
• Where should I work?
• What am I working on today?
• How long should I work?

For a full-time employee, these sets of questions have the same answer, every single day. Regardless of whether you’re feeling great or tired, regardless of what you have to do that day or what appointments you have, you have to answer these questions with the same location and the same time.

Using the same answer every day may be easy, but it doesn’t always fit. Sometimes you’re taking calls and networking all day; sometimes there’s a tropical hurricane heading past your state and it’s pouring buckets for the next week.

So, instead of a morning routine, have a morning algorithm. Here’s one I’m starting to use:

1. Pick out clothes the day before, based on weather and schedule.
2. If I have to take calls that day: pull out the laptop and work from home (It’s hard to take calls in a public space).
3. If there’s a meeting/event to go to: find a place to work near the meeting place
4. If it’s before X o’clock: Work at the faraway office (most productive for long blocks of time)
5. If it’s after X o’clock: Work at a nearby coffee shop.

With whatever algorithm you use, the instructions should be so simple, straightforward and brain dead that you can apply them in 5 seconds without questioning them. And once you’ve polished your algorithm over a few weeks of use, you should always follow it to the letter. Amend it with the greatest caution, because once you start changing it every morning, the benefits of reduced cognitive load disappear.

By doing all your thinking beforehand, you can save your brain cells for when your routine is over and your real work begins.

Lately there’s been a lot of privacy backlash over Facebook making a lot of information public by default. But one question that’s getting lost in the controversy deserves more attention. Why are they doing it?

I was lucky enough to be at a semantic web conference just after Facebook unleashed its new Graph API. What Facebook has done, for those who aren’t familiar, is changed Facebook interests and “likes” into links. So now, instead of having a list of movies and hobbies you like, your profile now links to pages for Cooking and The Godfather.

Cooking has more friends than you do

These pages are wrappers of the same information that was on Wikipedia, IMDB and elsewhere, but with likes and wall posts added. If you click on some of the Wikipedia links, you’ll get taken not to Wikipedia, but to another Facebook page that wraps it. Facebook may be using Wikipedia’s content, but the experience and the information is controlled by Facebook and stays on facebook.com.

Facebook isn’t interested in (just) becoming an encyclopedia of things, though. Facebook is interested in becoming an encyclopedia of you. All of your interests and likes are now linked, via FB, to wrapper pages that Facebook manages. Facebook is the centralized database that stores all that information.

This is how Facebook sees you. From F8 developer conference.

—–

Is having an open graph everywhere inevitable?

The four founders of Diaspora, in an appropriately indie band pose.

At the same time that Facebook was transforming their site into a database of everyone, a group of four NYU college students got a writeup from the New York Times. Their project, called Diaspora, was (is) to make your personal encyclopedia entry private, so you can control your information and how it gets accessed.

But while people donated nearly 200 thousand dollars to their project, and much ink was spilled over how much Facebook was now sharing about us, one might argue that the change to a public graph a la Facebook is inevitable. After all:

• Many, some at Facebook, some elsewhere, have argued that privacy is less important to members of Generation Y.
• Plus, having a shared social graph is clearly better than one where you can’t see any of the nodes, right? Once we see how useful it is to share the music we like and the news we’re interested in through the graph, we won’t want to turn back.

There are two forces that will decide this future: developers and you, the user.

Developers matter

You may have noticed recently that a huge number of sites – CNN, Pandora, the New York Times and others – have started spouting Facebook like buttons. Some other sites have included ways to login using Facebook itself, making you verify your identity by using your Facebook information.

Websites do this because it makes things easier. People are (arguably) more likely to login to CNN using Facebook than entering their email address, setting yet another password that could be forgotten, and checking their email for some confirmation link they have to click on. The added convenience makes it worth it to connect CNN to your real name and identity. For some people at least.

Web developers are the key to this, because we end up building the technology that decides if our sites are linked through Facebook or not. And, frankly: There are few good alternatives to Facebook.

For login, there are simply no sites that have the coverage of Facebook. We as developers could let you login to CNN through Google or Twitter, but allowing Facebook logins and using Facebook likes is a necessity. Or at least, a de facto standard.

Let’s say you wanted to make an alternative: You’d need to make sure that there’s an easy way for developers to incorporate it into their sites, because for a long time, you will be dealing with developers who have to put Facebook stuff on their website, and you’ll be their spare time project. If they can’t just drop it in, they won’t!

How will people react to Facebook in the long run?

As a Facebook user, the question that really matters is not privacy or the social graph. Instead: What’s in it for me? What do I get if I share all this information?

There isn’t that much value in sharing my movie preferences to a bunch of people, only some of whom are actually my friends. Twitter has proven that a market exists for conspicuous sharing – wide, out there, open sharing – but a lot of Facebook’s privacy woes come from incidental sharing – the “oops, I didn’t know that was public” type of sharing. One person came up with his own solution – all of his Facebook information is now public.

To borrow from the earlier 2000s: Some of us signed onto FB thinking it was LiveJournal, and it’s turned into MySpace. Those of us who thought that will move on.

Facebook’s main utility, for me and the one person I asked, is to see if my friends have updated their pages, and to upload and look at pictures. None of this has anything to do with the social graph, and until someone comes up with a killer app involving me sharing my links to fourteen different things, it’s not going to matter to me. I’ll just turn it off, and my Facebook page will be just another one of the many abandoned webpages I’ve made about myself.

I recently had the chance to compare an organization to an email.

The organization has had a number of meetings and made some useful decisions. They’ve put up a small site, but want to reorganize it and update the copy (it’s currently a shell with no visitors). This has been going on over approximately two months.

The email was from an email list about a startup competition:

On Friday night individuals pitch ideas for new ventures

Teams form around the best ideas and then work over the weekend to develop and launch a prototype on Sunday

——

Useful observations:

• In a startup, your limiting reagent is how fast you can build. If you have a fast engineering team, you can iterate quickly.
• Just because a product is minimal doesn’t mean it’s viable. The key criteria: Does the user have something to do?

My version: Build every one to throw away.

Fred Brooks’ classic The Mythical Man-Month made this observation developing a new piece of software – inevitably, many of the problems that will need to be solved won’t be known until you try to solve them. The first try will incorporate many feelings of should-have and would-have that will be fixed on the second try.

This is much like a journeyman craftsman building his or her first bench, or planting their first garden. The first try is a learning experience. There is no way around this, nor should there be – learning is often best by doing. The valuable asset is the experience of the builder, not the product.

But we don’t live in the Middle Ages, the age of guilds. We live in the post-Industrial Age. Factory owners don’t consider hoarding all the toys or cars or electronics they produce as an asset. Henry Ford’s genius was to redesign not just the golden eggs, but the machine that lays them.

As programmers, we are not working on a code base. We are working on a machine that produces code. The machine is made up of us, our experiences, the tools we build to make code (which can be made out of code themselves!). This is where results come from, and we should spend our time tuning this machine by producing more, not protecting what we have.

Does this mean that we should throw out old code? By no means! Old code is one of the most efficient resources we have for producing new code. But every process of manufacture in the past has been made into a more automatic and refined process. How could we consider ourselves any different?

Does this mean that we should code like crap? By no means! As was once said, “If you write the first one to throw out, you will end up throwing out the second one as well.” The point is, write good code, but be willing to write new code – the point is to make not good code, but a good code factory.

This is how bureaucracy works. When you first see this, if you’ve had any experience with office bureaucracy, you may think “that’s true!” followed by “that’s crazy” followed by “that’s life,” and then you may run away because you want to watch scrubs or family guy or something.

I don’t have a TV though, so I will now try to figure out how to fix this problem. In the process, I will probably accidentally prove that it’s good not to watch TV. But that’s ok, because I’ll probably watch TV after that, disproving everything I just said.

Bureaucracy

is all about the disconnect between people and their incentives. At big corporations and gov’t organizations, you don’t earn the money you spend; you get a budget. If you want to spend more, you use up all the money you have and then try to talk the right person into giving you more next year.

When one person runs a business, they try to do what’s in their best interests. When two people work together, they will tend to hold each other accountable. When one thousand people work together, the interests of the group become disconnected from the interests of the individual.

The Myth of the Rational Voter describes this very well. It may be true that how well my company does affects my well being, but I make a very small difference to a big company, and this is outweighed by even a small incentive to act in a different way.

Transparency

Once a group gets big enough, we must find a way to realign the incentives of its members with the incentives of the group.

One of the things we mentioned earlier was holding people accountable (in the case of small groups). We also noted that people are disconnected from their incentives in large groups. If we can make the individual’s accomplishments public within the group, we can measure them.

There is a weakness to this approach:

1. The individual is rewarded on appearing to be helpful, not actually being helpful.
2. Even if people are honest, there is no way to say for sure what is helpful–if the organization is driven by thousands of people, it may be impossible to accurately measure individual contributions.

The Free Market Approach

The most well-known of getting people to act in the general interest is market economics, which argues that people who are free to act in their own best interests and have a way of exchanging value with others will maximize the public’s well being as a whole. It’s interesting that though the economy is largely free market, corporations internally are run more like dictatorships; this may imply that there is some reason for the free market organizing principle to break down within a corporate body.

Is it, hypothetically, possible to change an organization into a group where everyone buys and sells from each other? If so, how would we preserve the advantages of a corporate structure?

For example, say I have a software product and I need someone to develop features for the latest version. Insofar as the product is modular, I can pick a module and contract out bids to people within my company. So far, so free. However, systems within a company tend towards natural monopoly–it is unlikely that any company would throw out its working team to replace it with a new, less experienced one.

Need that be absolute? Perhaps part of owning a process is having the ability to transfer knowledge about that process. If you can force the team working for you to make that process transferable, it changes from their asset to yours (and lest this seem unfair, you would likely pay for the privilege). The same goes for software as it does for factories or stores – if you don’t have the information to operate it with a different team, then you don’t own all of it.

Final Comments

But this is just one objection – an example amid many. The U.S. Army, certainly, could not operate along those principles. The power of choice is necessarily limited when it comes to a professional army.

Transparency can get people to operate in a way that appears good, though it wouldn’t actually make them act in the way best for the organization overall. There won’t always be two or more actors to compete for the same work and push the best one to the top.

Individuals can generally be counted on to act in a way that benefits them; the free market is largely based on this principal. Bureaucracy is typically about getting large groups to act in a way that benefits them as a whole; even when this is parallel to the individual example (large numbers of bureaucracies competing with each other) there is a necessary weakness–individuals are still individuals, and try as we might to get their interests to all point the same way, they will inevitably shift in their own direction.*

This may in fact be a good thing. Do we want the Mafia’s cronies to act with single minded purpose towards advancing the Mafia, or would we rather they skimmed and subverted?

That individuals can be counted on to act in their own interest more reliably than a group interest** may act as a bulwark against the power that a group has against an individual. Plenty of groups have turned against each other in history, but it is likely that there are injustices that have never occurred, simply because the unjust lacked the ability to act with sufficient unity.

* If you want to reduce this subversion, forcing people to act in the open is highly useful. Mobilizing the forces of conformity is likely to encourage people to act more subtly in their interests, however.

**(at least, when the group is not pitted directly against another group)

(originally started January 20, 2006 @ 00:22)

Ever since the RIAA crackdown on illegal downloading, there’s been quite a drop in CD sales. And iTunes sales have gone up substantially. But the reason isn’t that users feel that DRM is bad, just that illegal downloading and iTunes are both cheaper. The Kindle not selling enough? Maybe because it’s $400. [no citation available, just see for yourself] At least the books are cheap.

Perhaps, it’s not about restricting rights all the way to the point of controlling everything we see and do. Maybe it’s just about the money.

Bill Gates is my personal hero, which should come as no surprise to those who know me as a Kool-Aid drinking fanboy. However, the reason I admire Bill Gates has little to do with Microsoft itself. He exemplifies personal qualities that I admire, including technical breadth, ambition, and commitment to effective philanthropy. I’m sure Bill has made some mistakes over the years, as many would like to point out, but the way I see it, the best way to judge character is to see if a person really puts his money where his mouth is. Bill delivers, though I suppose that makes his mouth pretty big.

Adios, billg! So long, and thanks for all the fish/DOS/qbasic/etc.

-Tim