Selling AI to People

Let's say you're somewhere in the field of AI. You are somewhere from AI researcher (only uses their keyboard to check their mail, every 1 minute) to AI developer (fingers rarely leave keyboard). You know what AI is. AI is simply the automation of things, under various circumstances. There are certain laborious, dangerous, prohibitively difficult tasks, so we make software or machines to do them.

If you utter "Artificial Intelligence" to a non-technical person, a completely different image appears in their mind. It's something a bit more like this:

or worse, this:

I've actually had people ask me if machines will one day "love".

The non-technical person has a completely different image of what constitutes AI than you do. The consequence of this for the entrepreneur is, if you tell someone "My software is revolutionary because it uses Artificial Intelligence!!", they will undoubtedly become excited, and then become utterly disappointed when you actually show them your software.

No where have I learned this lesson more bluntly than when I was pitching my startup to a venture capitalist.

Brief story. This was back when our software was something completely different, and we were pitching it as basically an intelligent task management system. We walked into the meeting, and by this point, all the investor knew was it involved artificial intelligence, and he was very excited. As we were giving the presentation, you could slowly see the man deflate. When it finally came time to give our demo. He watched the system assign unfinished tasks to time slots, and exclaimed "it's just putting tasks where ever there's free time." I tried to explain that no, the system is actually sorting through many constraints, and figuring out what the possible orderings there are for tasks given the goals that have to be accomplished, but he was having none of it. "This isn't...artificial" he said in disappointment.

AI is about solving problems that are too difficult to do by hand. It enables you to solve extreme, prohibitively difficult problems. For example, finding content on the internet relative to your query involves looking at TRILLIONS of pages and examining them. One person could not do it in his lifetime, so Google uses AI to do it for you. But you'll notice Google does not ever say "We use AI! Come use our search engine!" They just solved a problem no one else could nearly as well, and let this fact speak for itself.

A similiar sentiment has been expressed recently by the infamous Ted Dziuba, one of the creators of Pressflip (formerly Persai), a new content recommendation site:

How Does it Work?

That's a good question. After dealing with angel investors, VCs, users, and anybody who isn't an engineer, the answer in my mind is, nobody gives a shit. Really, nobody cares about your algorithm or how revolutionary you think it is. All people care about is a system that shows them things they want to read.

I'm here to tell you, man, that is right on.

Your software is not revolutionary because it uses Artificial Intelligence. Your software is revolutionary because it is able to solve a problem that was until this moment not solved. *That* is what you must pitch.

Uhh....words go here.

Needless to say, I've been away cooking up good things at Method in Mind. We're just about to enter testing phase for Operatus and do a test install at a company. Just as a note, I've been averaging approximately 9-10 cups of coffee per day. Actually, the other day I went to the eye doctor shortly after getting the intravenous caffein delivery, and I freaked the shit out of the lady checking my blood pressure. She didn't quite breathlessly mouth the words "You...are going.....to DIE" at me, but....she was concerned.

Next time you're at a meeting, it may help to conjure this image in your mind:

First and last political post (hopefully)

UN says there is absolutely zero evidence that Iran has a nuclear weapons program.

Place a wet blanket upon your mania.

IE users can signup now......yay!

Okay, you can now sign up for the Operatus mailing list at Method in Mind if you're using IE.

fuuuuuuuuuuuuuuuuuuuuuudge

In the past few months I've exactly zero moments to write a post.  To serve as the vanguard post for many more to come, here is a silly you tube video:  

Method in Mind signup

Hey everyone! Sign up for the Method in Mind email list. It's the only way to enter to be a beta tester!

Cuong Do talks about YouTube's Scalability Architecture

Cuong Do, one of the original engineers at YouTube from the beginning, talks about how YouTube achieves scalability in a Google Tech Talk. Definitely very informative.

PS3 No good for AI, eh?

I'm a little late with this one, but perhaps you've seen this interview with an Ubisoft Splinter Cell developer. In the interview, he states that "the PS3 can't handle AI nearly as well as the 360."

First off, let me state that I am in no way a Sony fanboy. I don't even own a current generation console. That said, this statement is completely devoid of meaning.

You know what AI is? A plain old algorithm. A search process, selecting a rule from a rule base, or running through a pattern matcher. Saying the PS3 cannot do AI is like saying it cannot do sorting or print the first ten integers.

What some people are referring to to support the Ubisoft developer's argument is the PS3's relatively poor branch prediction ability. The argument then being, since most game AI is simply a giant switch statement, performance will suffer.

My response to this is, developer ignorance of AI is not a good excuse. If you think AI amounts to a giant switch statement then you need to go read a book, or a modern survey of AI planning.

On Advertising Based Business Models

When you've got a piece of software, and you're thinking about how to get money in exchange for it, there are clearly many sources advising you to give it away for free - the logic being that you monetize eyeballs. By far, the most popular way to do so is some form of advertising. One way this works is through site adds, where if you're clever you'll figure out how to make the adds targeted, such as Pandora. Another way this works is through some sort of affiliate program. In an affiliate program you get a fraction of the payment a customer gives a company you helped hook up through your site. Any coupon finding service works like this, like this one.

The typical value you get in exchange for an ad click, though, is extremely small. To get a feeling for the state of things, check out Guy Kawasaki's review of one year of blogging (for those unaware, Guy Kawasaki is an extremely popular blogger on tech entrepreneurship, who also happens to be a venture capitalist). He got on avg. 6,200 page views per day, and with the adds placed on his site, generated $3350 during the entire year. To contrast this, look at something like MySpace, which evidently makes $30 million in advertising dollars. And that's because they get, according to TechCrunch, 15 million daily unique logins.

What I"m trying to say is, this model makes financial sense only if you can attract a gigantic amount of traffic to your site. So much so that, in order to reach sufficiently many people, what you're doing has to appeal to the absolute lowest common denominator. Everyone has to be in the target market for your software.

Do you think Mathematica could survive with this model? Photoshop? Maya 3D? These are spectacular achievements of software that absolutely could not survive if the creators were gripped with the Web 2.0 mania that demands you give your software away.

So, if the circumstances are right, be unafraid to do the old fashioned thing and exchange what you've made for money. If you possess specialized knowledge which you've used to make something great, people will pay you for it.

Serious Action Tonight

Nuff said.