Facebook's Yann LeCunn Discusses Digital Companions and AI | Hacker News Reader

Facebook's Yann LeCunn Discusses Digital Companions and AI

bits.blogs.nytimes.com

58 pointsxenophon15 comments

I now think Facebook is going to give us Artificial Facebook friends. Modeled on your interactions with your Facebook contacts, combined with the data of a million other connections, it will be more attentive, always available, knows just how to cheer up your mood, or find you things that interest you when they are going viral. This may be many years out, but Zuckerberg needs to think about this future too.

Something fell into place for me when I read this interview. Like when Google bought YouTube, it took me a while to realize that Google would own internet video. Facebook is sitting on a huge pile of social data. With social graphs and tagged faces the more obvious ones, they also have a huge corpus of natural language and personal likes. With this one could create AI that is much more down to earth and friendly to communicate with. It may posses a form of emotional intelligence. It may have knowledge on social interactions, status, moods, human relationships etc. it knows what makes us tick.

Contrast with Google which basically indexed the entire internet. I expect an advanced future interaction with Google AI to be more factual. The kid you want to do your homework for you, not the kid you hang out with later, who suggests new Jazz songs for you to listen to when it starts to rain outside.

I think Google is going for the "Star Trek computer", void of a personality. Specifically an interface rather than a personality. Notice how Google's "assistant" is the only one without a name or persona, in contrast with Cortana, Siri, and whatever Facebook will surely create.

I think it's interesting that FB also could own relationships.

With their data they could (though they wouldn't for many reasons) accurately predict who is about to start a relationship with whom, who people are interested in, who is cheating on whom etc.

The cool things you could infer from the data they have available would be a lot of fun to play with.

Like a modern peeping tom. Good times!

I don't know how close this plays into what you are saying but maybe somebody will find it interesting.

In the summer of 2013 I was dating two women and I liked them both equally and had an extremely hard time deciding which one I liked better. I felt terrible about not being able to pick one and felt like I was hurting both of them. Facebook did nothing to help the situation by consistently plastering both of their pictures and links to their profiles all over every facebook page I visited. I know it had both of them marked as people that I interacted with in a different way than other people I was friends with. I've always wanted to know how deep and how many descriptive words and categories they have to identify the types of relationships people have or had. To this day when I type even one letter of any ex-girlfriend's name into the search bar they will always be at the top of my list. Facebook consistently eggs me on to check up on girls from my past and I really don't like that as I feel they know exactly what kind of emotions they could be stirring up.

Maybe I'm being a bit ridiculous about the entire thing but it's something I've always noticed and wondered if other people had opinions on.

Facebook has already managed to turn a large portion of the population into what I would consider "artificial", so may as well go one step further and make it official. I can't wait to have a group of artificial friends to tell me how great I am and report everything I say in return back to Facebook!

I had a similar reaction. Many of us who spend a lot of time online interact frequently with people we've never met as it is. In 3-5 years, we could have digital followers on social networks who are indistinguishable from real people.

Imagine that they're more attractive, more flattering and more interesting - what happens next?

I can see us becoming like the animal that dry humps some false approximation of its mate, unable to tell that it's stepped away from reality and is just fulfilling its desires without procreation.

Great comment

> unable to tell...

Really? Might it be that we purify memetic procreation of its inefficient dependency on support systems engineered for DNA-genetic procreation?

>Imagine that they're more attractive, more flattering and more interesting - what happens next?

Nothing, really. A voice in a chatbox is still only in a chatbox, and this whole prospect will creep people out sufficiently to make them leave Facebook.

(My mother apparently has something called "Google Cards" on her phone that creeped her out by reading her email and giving her reminders of things. Facebook false-friends that imitate living people without being living people are going to creep her out even more.)

On the other hand, the scientist in me notes that if you're actually trying to make my Facebook False Friend more interesting than a real person, you're going to have to teach it to fluently and accurately discuss research-level science and mathematics, teach me analytic philosophy, and then argue about which anime characters are best. That's really asking quite a lot! I don't think Facebook will accomplish anything quite that "scifi" in the next few years.

You're judging a product now. I'm telling you to judge it in a few years when it is more convincing, more subtle, etc. People already "fall in love" online with people they've never met, partly because someone showed some interest in them. Many are scammed easily online because their greed has them look past telltale signs.

You've just interacted with me on some level, but what's to say I exist? Pretend I only have social network profiles online for you to background check. Those would be easy to dynamically and convincingly generate.

And don't just imagine your reaction, but the reaction of multitudes of people who aren't skeptical about religion, about crystal healing, who are superstitious, who watch whatever is put in front of them, etc.

I bet someone could scan my body of online work and design an array of profiles that might pique my interest.

The hard part isn't generating the profiles -- that's dead easy. It's making the programs generate fluent conversation in natural language with interesting domain knowledge -- that's still believed to be AI-complete.

Half the population struggles with fluent conversation!

> When you are playing chess, at some point you make a mistake, you may go back several steps that were “correct” to find the one that was wrong. When you fall off a bicycle, you think of when you lost your balance. Deep learning does that. The credit assignment in a deep learning exercise can be tens, even hundreds, of levels deep.

This sounds like reinforcement learning. Anyone know what he's talking about, some RNN with 'tens, even hundreds' of levels of feedback that hasn't died away?

Pretty sure he's just referring to backprop[0] correcting weights more dramatically at one layer vs. another layer

[0]http://en.m.wikipedia.org/wiki/Backpropagation

Reply on news.ycombinator.com