Copyright Google Inc.
Last week, with much fanfare, Google revealed it’s latest party-trick, “Duplex” – an upgrade for their virtual assistant.
At the launch, Google CEO Sundar Pichai announced: “… our mission for the Assistant is to help you get things done”.
No controversy there, you might think.
He continues; “it turns out, a big part of getting things done is making a phone call”. Again, on the face of it – you could have no issue with this, but then he carries on: “we want to connect users to businesses in a good way”.
All very laudable, Sundar – but here’s the punchline: “businesses actually rely a lot on this, but even in the US 60% of small businesses don’t have an online booking system setup. We think AI can help with this problem…”.
What proceeds is an incredible demo of Google Duplex calling first a hair salon and then subsequently a restaurant and making a booking on behalf of the user. Here’s a link to the video, it’s well worth watching:
For those of you who don’t have time to watch the video, what transpires is a very human sounding synthetic voice navigating a complex set of questions with the two businesses in a flawless manner.
Not only did it understand the context of questions and statements throughout the thread of a conversation (at time of writing in Q2 2018 – any user of Siri, Cortana or Alexa will appreciate the novelty of this), but it also introduced casual interjections and disfluencies such as “um”, “Mm-hmmm”, and “er” into the speech.
These are, of course, impressive achievements, and while the audience at Google IO 2018 cheered on the demonstration; what struck many commentators including me is how desperately Google and other tech companies need a lesson in ethics lest these demos backfire.
What most people clung on to is how Duplex did not announce itself as an AI at the start of the call. In fact, the tonal inflexions and casual speech feel like they have been deliberately designed to mask the fact that the two businesses were, in fact, speaking to a machine.
There are two issues with this. Firstly, there is a long-established research science principle that informed consent is required before conducting experiments with human subjects. Facebook famously breached this in 2014 in a partnership with Cornell University when studying ‘emotional contagion’ – essentially whether a perceptible difference could be seen in the mood of Facebook users when exposed to a news feed that was more positive or negative in tone. One would have hoped that the folk in Menlo Park and neighbouring Cupertino might have learned a lesson from this.
The second issue – the bigger one, in fact, is to what extent should we allow machines to sound like people – how lifelike ought the voice be? And should AI’s be required to announce themselves in conversation? These are not questions I have answers to, but what is critical is that we engage in this discussion in schools, universities, coffee shops and at dinner tables – we need to set the rules of engagement with our future relationship with machines before those rules are dictated to us by the tech industry.
To me though, the most serious issue; and the one which it seems all commentators on the subject have failed to spot is the duplicity of Duplex. Looking back to Sundar’s commentary, the narrative that was built for the demo was that their intention in the creation of Duplex was to help people get things done, and to help connect businesses and users in a ‘good way’.
If this was indeed the starting point for their R&D, then when did the decision get made to create a voice-bot that pretended to be human and essentially wasted the time of those two companies (one presumes that Lisa didn’t actually want a hair-cut, nor did a party of four visit the restaurant that was called)?
The issue that Sundar himself recognises is that 3 out of 5 small businesses don’t have an online booking system. He’s right when he says that AI can help, but that’s the problem that Google should have solved. How to enable small companies to take advantage of AI and become more competitive in the market vs those companies with greater capital and technical wherewithal to invest in Artificial Intelligence.
By introducing Duplex into the hands of users, what Google actually achieves is to further reinforce the uncompetitiveness of these small Enterprises. You could imagine SMEs bombarded with phone call after phone call from smart AI systems like Duplex. This is not solving the problem in a good way.
Finally, one can imagine that hair salons and restaurants are not the markets where this technology will proliferate. Today automated phone diallers cause misery to many who are on the receiving end of scammers, debt collectors and unsolicited sales calls. It’s these firms that would most seek to benefit from Duplex.
The technology industry is fast reaching the point where it is no longer able to regulate its own activities for the benefit and safeguard of those with limited economic power. We might be able to avoid regulation, but only through societal wide dialogue on the subject. I fear, that once Pandora’s Box is opened, it will be almost impossible to curtail the limits of technology.
We have a golden opportunity with AI to design the future that we want to have. Let’s not waste it.
Interested in the effects of technology on society?
Copyright Google Inc.
Last week, with much fanfare, Google revealed it’s latest party-trick, “Duplex” – an upgrade for their virtual assistant.
At the launch, Google CEO Sundar Pichai announced: “… our mission for the Assistant is to help you get things done”.
No controversy there, you might think.
He continues; “it turns out, a big part of getting things done is making a phone call”. Again, on the face of it – you could have no issue with this, but then he carries on: “we want to connect users to businesses in a good way”.
All very laudable, Sundar – but here’s the punchline: “businesses actually rely a lot on this, but even in the US 60% of small businesses don’t have an online booking system setup. We think AI can help with this problem…”.
What proceeds is an incredible demo of Google Duplex calling first a hair salon and then subsequently a restaurant and making a booking on behalf of the user. Here’s a link to the video, it’s well worth watching:
For those of you who don’t have time to watch the video, what transpires is a very human sounding synthetic voice navigating a complex set of questions with the two businesses in a flawless manner.
Not only did it understand the context of questions and statements throughout the thread of a conversation (at time of writing in Q2 2018 – any user of Siri, Cortana or Alexa will appreciate the novelty of this), but it also introduced casual interjections and disfluencies such as “um”, “Mm-hmmm”, and “er” into the speech.
These are, of course, impressive achievements, and while the audience at Google IO 2018 cheered on the demonstration; what struck many commentators including me is how desperately Google and other tech companies need a lesson in ethics lest these demos backfire.
What most people clung on to is how Duplex did not announce itself as an AI at the start of the call. In fact, the tonal inflexions and casual speech feel like they have been deliberately designed to mask the fact that the two businesses were, in fact, speaking to a machine.
There are two issues with this. Firstly, there is a long-established research science principle that informed consent is required before conducting experiments with human subjects. Facebook famously breached this in 2014 in a partnership with Cornell University when studying ‘emotional contagion’ – essentially whether a perceptible difference could be seen in the mood of Facebook users when exposed to a news feed that was more positive or negative in tone. One would have hoped that the folk in Menlo Park and neighbouring Cupertino might have learned a lesson from this.
The second issue – the bigger one, in fact, is to what extent should we allow machines to sound like people – how lifelike ought the voice be? And should AI’s be required to announce themselves in conversation? These are not questions I have answers to, but what is critical is that we engage in this discussion in schools, universities, coffee shops and at dinner tables – we need to set the rules of engagement with our future relationship with machines before those rules are dictated to us by the tech industry.
To me though, the most serious issue; and the one which it seems all commentators on the subject have failed to spot is the duplicity of Duplex. Looking back to Sundar’s commentary, the narrative that was built for the demo was that their intention in the creation of Duplex was to help people get things done, and to help connect businesses and users in a ‘good way’.
If this was indeed the starting point for their R&D, then when did the decision get made to create a voice-bot that pretended to be human and essentially wasted the time of those two companies (one presumes that Lisa didn’t actually want a hair-cut, nor did a party of four visit the restaurant that was called)?
The issue that Sundar himself recognises is that 3 out of 5 small businesses don’t have an online booking system. He’s right when he says that AI can help, but that’s the problem that Google should have solved. How to enable small companies to take advantage of AI and become more competitive in the market vs those companies with greater capital and technical wherewithal to invest in Artificial Intelligence.
By introducing Duplex into the hands of users, what Google actually achieves is to further reinforce the uncompetitiveness of these small Enterprises. You could imagine SMEs bombarded with phone call after phone call from smart AI systems like Duplex. This is not solving the problem in a good way.
Finally, one can imagine that hair salons and restaurants are not the markets where this technology will proliferate. Today automated phone diallers cause misery to many who are on the receiving end of scammers, debt collectors and unsolicited sales calls. It’s these firms that would most seek to benefit from Duplex.
The technology industry is fast reaching the point where it is no longer able to regulate its own activities for the benefit and safeguard of those with limited economic power. We might be able to avoid regulation, but only through societal wide dialogue on the subject. I fear, that once Pandora’s Box is opened, it will be almost impossible to curtail the limits of technology.
We have a golden opportunity with AI to design the future that we want to have. Let’s not waste it.
