A new Google patent and a Turing test for social media

Here is the latest:

It’s unlikely the majority of us are so overwhelmed with tweets, Facebook posts, emails and texts that we need someone – or something – to reply on our behalf.

However, this hasn’t stopped Google filing a patent for a system that would do just that.

According to the details of the latest patent filed by a software engineer at the firm, Google’s automated system would work like a social media bot and submit posts on a user’s behalf.

It would do this by scanning that user’s previous posts and replies on sites such as Facebook, Google+ and LinkedIn, as well as their emails and texts messages, to learn how that person writes.

The bot would then write replies and responses to future posts in a way that mimics that person’s usual language and tone.

The more a person uses the system, the more the bot can learn the type of responses they write and this would make the suggestions sound more human and realistic.

The full story is here.  Some of you will recall that I discuss this possibility in Average is Over, and consider an equilibrium where many people secede from email altogether and the value of face time rises.

For the pointer I thank the excellent Mark Thorson.

Comments

'and consider an equilibrium where many people secede from email altogether and the value of face time rises'

A true 'Thus Spake Friedman' moment.

Not much different from being a prominent academic and having grad students ghost write. Just a bit more automated.

Former academic here--grad students don't ghost write for prominent academics. Yes, there are cases of professors stealing grad students' work, but it isn't "ghost writing" if we define that as a sanctioned activity.

If anything it's the other way around. Grad student is first on the author list, but the prof wrote the text.

Ok, so I admit I saw this app on Twitter a few weeks ago and passed it to my FB friends ... setting aside "privacy" concerns about FB content ... it is great fun. Here was the HT on Twitter: https://twitter.com/lydiadepillis/status/401719797928316930 Yes, note the ClaudiaBot is even a bit odder than the Claudia MR commenter. What was interesting was from the friends who played along I could have picked out their bots even if unsigned.

I do not see why this is a disconcerting or at least a surprising trend ... especially if played freestyle. I would just as soon have my bot handle meeting schedules, help me follow simple daily tasks, etc. Computers are stupid, they obey rules (granted some are complex) it's not a bad compliment for many of us, who struggle with rules. I don't think this trend will necessarily elevate in-person contact but it will make human-to-human contact more precious, if you can get it or still want it. Of course, to prove you're not your bot you may have to be odder than your bot.

Sounds like something companies might pay for since most small business owners have no time to post (much less the ebility to post in a way to attract their terget market).

A reincarnation of Alice / Eliza the ancient chat bots with a few tweaks on training?

Kibo is prior art. Automated replies was performance art back in Usenet days. Wikipedia can be your friend, if you didn't see it happening at the time.

I think you will

Never see

So wonderful

As a Tweet from me,

Watson

Google has been trying and failing to move into social media for years.

What would wide-spread adoption and implementation look like? Would the end result be a bunch of computers commenting, posting, and replying to each other without any human participation?

You could have flame wars and the humans involved not even know that they are flaming each other online.

What if you're online personality is that of a boorish troll? Would this bot go around posting mean comments on all your friend's posts and tweets on your behalf?

Charlie: People so thoughtless as to be compellingly mimicked ought not be in your newsfeed.

Friendbot: I know right! LOL

Black Mirror has a great episode that tackles this concept.

http://www.youtube.com/watch?v=ld9m8Xrpko0

Great. That's what we need: a bot to repeat "absolutely" and "So, ..." ad nauseam and bias N-gram with the world's worst grammar.

That device is in common use, it is called the database query. All google did was make it illegal to use their favored form of database query. But the entire market can dump the semantic of 'post' and adopt the semantic of 'query' and defeat both the patent and google's entire line of 'posting' business. Pursuing license fees would destroy their business.

What I want: a bot that filters out lower quality tweets and blog posts from people whose content is highly uneven.

I want a bot that summarizes blog comments into two or three recurring competing points of view.

Brad DeL did a related thing early this year: http://delong.typepad.com/sdj/2013/04/david-graeber-april-fools-day-post.html

What % of MR commenters are human?

The quality of spam is remarkably low in today's world.

That should be easy. I have a standard comment which fits 99% of all Facebook postings and almost as many blogs. It is "And elsewhere, children are starving."

Comments for this post are closed