Voice assistants such as Amazon's Alexa or chatbots, which are offered on many websites, are now invading the daily lives of more and more people. Meanwhile, computer programs can even automatically create texts or post messages on social networks. But what is behind it and how does it work?
In his 167-page booklet, free author Christoph Drösser, who lives in San Francisco, explains the approaches taken with voice assistants, chatbots or similar programs. This starts with a (dispensable) historical demolition of the 1769 ‘Chess Turk’ via the ‘Intelligence Test for Machines’ devised by Alan Turing (whose theories have tormented generations of aspiring computer scientists, such as the author of this review), and extends to the technologies used in today’s voice assistants such as Amazon Alexa, Microsoft’s Cortana or Google’s ‘Google Assistant’ (Google Now).
Drösser also does not spare any annotations, and also mentions the attempt with a chatbot that went quite wrong at Microsoft (see my blog post Microsoft Tay chatbot – partially failed experiment). A large part of the booklet presents interviews with people who work or research in the field of chatbots, language assistants and writing bots in the USA or who were active there until recently.
A separate chapter discusses the limitations but also the possibilities of statistical language analysis. In addition, the book deals with “the computer as an author”, i.e. machine-generated texts that caused a sensation in the last months of 2020 (see, for example, the Text generator GPT-3).
On the almost 160 pages, the author can only give an overview of the state of the art, whereby this information is outdated again in one or two years. I like some assessments regarding certain voice assistant technologies as well as the information that the voice assistants Alexa (Amazon) and Google Assistant (Google) perform better than Siri (Apple) in tests today.
Drösser also briefly discusses the risks of certain technologies. He explicitly mentions on four pages the risks that voice assistants, for example, listen to the communication of their owners and transmit it to the servers of the companies that provide these assistants. The author also discusses alternatives such as the open source project Almond, a virtual assistant being developed at Standford University.
What I personally lack is the assessment that many of these products, especially in the consumer sector, fall out of support or can be hacked within a very short time. The following links give a brief outline of blog posts that I have published in recent months on these topics from the perspective of a practitioner.
Cortana support dies for Android & iOS
Unattractive: Harman Kardon Invoke soon expensive e-waste?
Bundestag report warns parents about Amazon's Alexa
Amazon's Alexa cracked, numerous security vulnerabilities
Voice assistants as a hacking risk for mobile devices
Hacking voice assistants, the undetected danger
If you want to get a first overview of the topic and the philosophical questions regarding the use of these technologies, you will find a good overview in the Drösser paperback. The plant is certainly not intended for a deeper introduction to the individual areas of expertise. This also applies to readers who are hoping for practical advice regarding the use of voice assistants, or want to know how to recognize texts from bots in social networks or on the Internet.
Christoph Drösser: When things talk to us
From voice assistants, dense computers and social bots
160 pages, paperback, Duden-Verlag
1st edition (12 October 2020)
Price: EUR 16,00 Euro
ISBN-10: 3411742259
ISBN-13: 978-3411742257
Flap text: We talk to voice assistants and chatbots, read reports written by a computer, and let opinion bots manipulate us on social media. As a communication partner, machines play a role that was previously reserved for humans. We automatically impute consciousness and emotions to them. Therefore, the new language technology is more than a new interface. Christoph Drösser explains how it works and explores the possibilities, opportunities and dangers of intelligent language systems.