The Virtual Assistant Goes Elementary With IBM’s Sherlock
Source: Amir Mizroch
The Sherlock virtual assistant, a project of IBM IBM +0.69% and Cardiff University, is meant to work on the Apple AAPL +0.73% Watch as well as phones.
Cardiff University
A team at IBM and Cardiff University in Wales is working on a virtual assistant that isn’t part of the artificial-intelligence in-crowd led by the likes of Apple’s Siri, Microsoft MSFT +2.15%’s Cortana and Amazon’s Alexa.
While those apps crunch ever-larger amounts of data to do things like call someone from your contact list, tell you when it is time to leave for the airport or just have a witty conversation with you, the IBM/Cardiff team’s experimental Sherlock assistant takes a different approach. It gets its information not from IBM’s vast computer networks but from other Sherlock users, in what is called a crowdsourced model.
So unlike Siri, for instance, which works by learning to search the Internet or Apple’s own databases to understand your speech and look for answers to your questions, Sherlock (an acronym for Simple Human Experiment Regarding Locally Observed Collective Knowledge) gets answers to questions from the knowledge provided by other Sherlock users.
The Cardiff team says Sherlock is especially useful in emergencies, policing or even large events, where users can feed it information about things they see and hear.
For instance, if a user wanted to find the shortest line into a large music festival, the user could ask Sherlock by texting it: “What is the shortest line into the venue?”
Other users of Sherlock who have GPS on their devices at entrances to the festival would then be pinged with the question: “What is the length of your line?” to which they would text their approximation of the length of the line. Sherlock would then process that combined information and come back to the original user through a text message with a “best guess” as to the shortest line, said Prof. Alun Preece of Cardiff University’s School of Computer Science & Informatics, who led the team developing Sherlock.
At this point in its development, Sherlock interacts with people through text. One experimental version does use speech using Google 'sGOOGL +2.56% Voice interface. A smartwatch version uses Apple’s dictation interface.
Once complete and deployed, Sherlock could bring virtual assistants to developing markets like Africa, where texting is still the predominant communication platform for sharing information, and where there also isn’t always the mobile-broadband bandwidth or smartphone penetration to use data-heavy programs like Siri. And because Sherlock uses text, it is easier for the program to understand what the human user is asking it, as opposed to voice commands, which could be affected by intonation, accent or surrounding noise.
While not entirely based on artificial-intelligence computing, as Siri and Cortana are, Sherlock has some elements of computer self-learning, which it needs to accurately recognize and act on the text messages that users type in.
Still, its emergence is part of a growing wave of smart virtual assistants using large amounts of real-time data to help people make decisions and carry out tasks.
Apart from Apple, Microsoft and Amazon, the Google App also acts as a virtual assistant. Slack, a San Francisco-based team-communication app, is also looking to boost its virtual assistant, called Slackbot, a software program that mines internal company communications to answer questions and carry out tasks.
Facebook FB +1.23% recently rolled out a new virtual assistant called M, which also works through text messages. It isn’t only the tech giants that are making virtual assistants. New York-based startup X.ai this year raised $9.2 million for its assistant called Amy, whose task is to help users schedule meetings. Jerusalem-based OrCam this year unveiled Casie, a digital personal assistant you can clip on your shirt.
Sherlock uses a combination of local storage on the device and some remote server processing. It can keep working even when there is no connection to the cloud, because in some circumstances it can fall back on what is stored on users’ phones. “That makes it really handy in front-line situations when the network is swamped or has gone down. It also works at big public festivals where bandwidth is often really limited,” Preece said.
IBM has been working on artificial intelligence for decades, mainly through its Watson program. On Sherlock, IBM Global Labs spokesman Jonathan Batty said, “While the technology is interesting and has potential, it is too early for us to comment about its future.”
| }
|