Thank you for Duplex, Google

With its experimental voice-based digital assistant system for booking appointments called “Duplex”, Google has created quite a stir. Some people are amazed. Some people are worried. Some are both. Oddly, I am neither. But there are couple of interesting take-aways.

1. Smells like Vaporware

Duplex and the demo shown are a perfect way to wow a geeky audience at a Google developer conference (mission accomplished). But for a company such as Google, is also easy to prepare the technology for a highly predictable use case such as booking an appointment. How natural and human-like would Duplex sound if the person on the other end of the line suddenly asks “Hey by the way, what do you think about the situation in the Middle East/the Eurovision Song Contest/Paleo?”. Exactly, these questions would not be asked during this type of call. Therefore, what remains is a highly trivial bot conversation, a well-functioning yet not revolutionary speech recognition system, and an indeed noteworthy human-like computer-generated voice. That’s for sure an achievement. But it does not say anything about the feasibility of the approach in more wide-ranging conversations, and neither about the actually potential of Duplex on the market.

2. “The new can’t do new things in *old* ways.”

The former Windows president Steven Sinofsky coined the adage that new cannot do new things in old ways. Yet this is exactly what Google does with Duplex. The investor Bijan Sabet puts it trechantly: “It reminds me of those apps that allow you to send a fax from a smartphone. It’s duct tape for old Infrastructure”

In fact, the whole scenario of people having to call to make an appointment should not exist at all (and probably won’t in a few years). Traditional online booking systems can do this better, as can chat bots. It’s strange that Google would try to solve such a transient problem over actual innovating in this field. Sabet again:

“Instead I would rather see Google use it’s financial and engineering might to get everyone online and connected in a decentralized way via API so office hours, reservations, communications arent addressed thru computers pretending to be humans.”

But when complaining about the misguided allocation of resources and lack of innovative thinking in regards to Duplex, the premise is that Google actually would be serious with this. But the company’s actual intentions might be very different…

3. Google kicked off an important debate

Considering the previously mentioned points, it’s unlikely that the Google management presented Duplex with the sole (or prime) intention of actually launching such a service. More probably at least is that, aside from the goal to entertain the Google I/O crowd, Google wanted to spark a debate and show how far technology has come to imitate humans in narrow, highly predictable types of conversations. Mission accomplished, again. I saw people discussing Duplex who otherwise never talk about cutting-edge tech topics. It’s important that even the non-techy crowd starts thinking about AI and its possibilities as well as challenges, because it is upon us and will affect everyone. We should be thankful to Google for doing this.

4. Our human discomfort with being presented with our lack of sophistication

On Hacker News, a commentator made this thought-provoking remark:

“People who answer phones to take bookings perform an extremely limited set of questions and responses, that’s why they can even be replaced by dumb voice response systems in many cases. In these cases, the human being answering the phone is themselves acting like a bot following a repetitive script.”

Indeed, part of the outrage and criticism in regards to Duplex could be caused by the natural human discomfort with being confronted with our own bot-likeness. Two years ago, I wrote a blog post pointing out that Twitter makes humans look like bots. Since then, I am noticing bot-like behavior everywhere.

Digital technology and particularly AI is challenging humans in ways which are scary, for the simple reason that it shows us our own shortcomings and lack of sophistication, such as when doing scripted conversations on the phone while insisting in wanting to keep the exclusive right to perform those conversations.

Machines will keep entering more and more territories of everyday life. The natural response for humans must be to actually focus on the areas in which we are and will be, for a long time, more capable than computers. And it also needs to be the realization that if we don’t want to be outperformed by bots, we must stop behaving like bots ourselves.

=======
Sign up for the meshedsociety weekly email, loaded with great things to read about the digital world. Sent to almost 600 subscribers (May 2018).
=======

Twitter makes humans look like bots

The advances in artificial intelligence and the rise of businesses that develop and employ chat-based bots mean that it gets increasingly hard to know whether you are dealing with a machine or a human being. The technology behind bots has gotten so sophisticated that it can require a longer conversation in order to be sure that there is no person of flesh and blood on the other end, as illustrated in this exchange with a Google support “employee” which I linked to in yesterday’s reading list. Basically, one has to run somewhat of a freestyle Turing Test. In other cases, the opposite happens: Users assume they are interacting with a machine, but in fact are having a chat with a real-human who only pretends to be a bot. An “Anti-Turing-Test”, as conducted in this example with Facebook’s experimental personal assistant M, can reveal this.

Bots pretending to be humans, humans pretending to be bots – sounds a bit bizarre, doesn’t it? Here is something else bizarre:

Think about what’s typical for a contemporary chatbot, those being used by large companies for customer service such as in the example above by Google, or by telecom operators (I recently had a chat interaction with T-Mobile which made me suspicious that I was conversing with a machine pretending to be a human): Continue Reading

AI-powered chatbots and the future of language learning

Chatbots are one of the next big things. Facebook’s experimental feature M, Slack’s Slackbot or bot-add ons for existing messengers like Whatsapp and Telegram such as Murdoch or WhatsBot show where we are heading: Into a world of conversational interfaces based on texting. Despite all fancy interaction tech available, texting has evolved as the world population’s most favourite way of interaction. Over the next years, more people will start to have text conversations with machines, so called bots. Conversations that thanks to advances in artificial intelligence and machine learning could feel pretty much like those with people from our “human” contact list. Even though sometimes still, what is presented to be a bot actually could be a human.

One area in which I hope that text-based conversational interfaces will flourish is language learning. The other day, I myself acted as a language learning bot, and the results were promising. Continue Reading