You wanted innovation? We gave you Clippy the Paperclip in your IM client
The Bot Wars are proof the Valley is bust
Comment Look no further for proof that Silicon Valley is now running on its own exhaust fumes than the latest hype: The War of the Clippys.
A decade after Microsoft banished its helpful cartoon “Assistant” Clippy from Microsoft Office, Clippys are popping up everywhere, like a plague of rodents. Don’t laugh: Clippy is now a “platform”, with a Clippy API. It's even a vision thing. Yesterday, Google followed Microsoft and Facebook in squirting a bot assistant into a new IM chat client, the umpteenth attempt in a long line of failed Google IM chat clients.
Three things are wrong with this, and they’re all things that the platforms may yet regret.
Google is using its AI smarts to create a messenger that handles your conversations for you https://t.co/4PBmGBHB76— Jack Clark (@jackclarkSF) May 18, 2016
Probably the weirdest thing about Google Allo is how it relies on seq2seq learning, which is very, very new. https://t.co/eytpx6HVpq— Jack Clark (@jackclarkSF) May 18, 2016
Firstly unless your job depends on it, the chat bots don't show how AI research is going to take your job, or usher in a new age of robots. All that this purportedly Earth-shattering, bleeding edge can do is prompt you for a response when a friend says “Hi” – autoprompting will it be “Hi” or “How are you doing?” although it may “learn” “Alright then yer f*cker”, and spam you with local commercial offerings, no differently to intrusive local search today.
Behind the hype lies the fact that AI research hasn’t made any serious breakthroughs in recent years; computers have got faster, so the same brute force techniques now deliver their results more rapidly. They’re just as stupid, but faster.
Apologies for repeating it, but Jaron Lanier's observation that in order to make AI seem impressive, we first have to make ourselves stupid, holds true. My own personal litmus test for AI breakthrough is whether a computer can distinguish between sarcasm and irony. So far, none can. Can the AI actually do humour? Please peruse the many witty entries to last week’s Spotify "name-your-own-Alan-Partridge-style TV show competition. Now tell me how close Google or Microsoft or Facebook’s AI comes even remotely close to passing the Turing Test.
Wake me up when an AI bot can do funny as well as Reg "Commentards"
By way of comparison:
Stuck for words? Don't expect anything original from the AI
If the bots are the best Silicon Valley can do, then it’s poor reflection on the Valley. And it’s actually a pretty good argument for closing down publicly funded AI research, and putting the money into something more useful, like… I dunno, cheap energy, antibiotics resistence … or pretty much anything. Not that I am advocating that, but somebody will, and quite persuasively.
The second problem is that it shows the limitations of the Silicon Valley world view. It has been pointed out that almost every idea to be advanced in the past decade as “innovative” or “disruptive” seems to spring from the mind of a socially inept and self-centered 14-year-old adolescent male. That explains all the concierge services. Now it explains why Silicon Valley companies need to invent an imaginary chat companion. Nobody else will talk to the buggers.
I see a third problem, and this one could be fatal. When an AI inserts itself into your private conversation, what that means is that your private conversation reaches a lot more computers than you can imagine. Previously the only computers involved were doing dumb transmission. Now they’re actively processing these private exchanges, second-guessing your emotional state. Bear in mind too that the "beacons" which Silicon Valley wants us to install in our homes (at our own expense) never delete anything. Google never deletes your interactions with Google Now - storing every second of your voice calls forever. Google argues that this does not compromise your privacy, because it “anonymizes” the recordings if you ask nicely or delete the account.
Like sterilsed milk. Open it, and it isn't > Posting 'anonymized' research data may pose threats to patient privacy https://t.co/2cENZDea1a— Guy Herbert (@guy_herbert) May 18, 2016
It’s still gradually acquired a library of containing an acoustic signature of every citizen’s voice. Priceless.
What the chatbot wars reveal, then, is a Silicon Valley that, far from being disruptive and innovative, is desperately uninventive, and creepy with it. Now that the low-hanging fruit (search, free music, free email) has been plucked, we can see them for what they are, quite transparently, which is consumer data slurping operations. Only by putting a real digital economy in place will the next generation of nerds be incentivised not to treat huge aggregations of personal data as an asset. I hope you’ve got some more ideas. ®