Machine Sirens and Vocal
Intelligence
Luciana Parisi
After Siri, the virtual assistant from Apple, a new-generation speech
recognition software platform called Viv will soon come to unite services and
devices into one unbroken vocal activity.1 Viv will connect intelligences
across services and users in order to offer immediate resolution between
query and delivery. With Viv, we will enter the realm of artificial
intelligences programmed to have a conversation with us. This does not just
involve continuous interactive feedback, but seems to realise Gordon Pask’s
imagination of a machine that can initiate a dialogue,2 find out what we like,
and offer us alternatives that we had not thought of.
Writing and visual interfaces will be enhanced with automated oral
communication, which replaces hand-to-eye with ear-to-image correlation.
Oral automation thus promises a synthetic time, replacing the steps of writing
and self-reflection with the speed of sonic wisdom, emitting inhuman
frequencies that will prove irrevocably alluring to us. Whilst speaking with
aural bots is still frustratingly limited to exchanging a set of utterances such
as those emitted by automated marketing bots or service providers, research
in AI aims to replicate the pre-alphabetic stage of uninterrupted transmission,
where speech-to-speech communication formed the basis of thinking aloud
(i.e. before thought could be formalized). It is only because the vocal
constituents of speech could be mechanized and recorded that technology
became embedded in social thinking, not only relativizing physical distance
but also giving rise to artificial intelligence systems that could no longer be
perceived as mere instruments. Whilst Turing’s paper ‘Computing Machinery
and Intelligence’3 referred to text-based conversations that would supposedly
determine whether or not a machine could think like a human, the synthetic
voice of intelligent assistants today rather shows that thinking involves not
just a sequential arrangement of symbols (as if these were hardwired to the
brain), but must include cognitive levels of affective communication. It has
thus been revealed that intelligence has a sonic architecture of implicit
wisdom that works not through deductive inference, or the logical
conforming of results to premises. If Skynet-Capital is increasingly investing
in synthetic voice intelligent interfaces, it is because it seeks the sonic
unification of products, services, and users. As one commentator has already
anticipated, the oral intelligence of Viv will radically shift the economics of
the internet. Since it will simultaneously process unprecedented volumes of
data, its web portals will bring together information from diverse sources
allowing every service and business on the internet to become vocally
accessible. Viv’s masterplan is to become necessarily continuous, from
making a restaurant reservation to ordering a taxi and buying theatre tickets
in one unbroken conversation. Vocal intelligence will not simply avoid the
consequential temporalities of writing, but aims to surpass the speed of
typing, searching, and clicking. The gold rush for the next generation of vocal
intelligence is already heightening competition between AI giants Google,
Microsoft, Apple, Amazon, and Facebook for the conquest of the most varied
and complex aural interface.
Vocal intelligence announces the dawn of a post-internet era, leaving
behind the neoliberal image of network highways, now replaced with
situations-inclined programming, where agents write their own instructions
each time broadly diverse services connect together. Instead of having
responses already scripted by a programmer, as is still the case with Siri, for
instance, the new generation of virtual personal assistants is meant to learn
from queries about situations where there is not much specific information,
adapting to possible rather than already existing preferences. In short, AIs
like Viv provide a highly personalized service to users where
recommendations are offered in the form of conversation, raising interest and
maintaining a sophisticated level of dialogue. Not too dissimilar from Spike
Jonze’s depiction of the Service Provider Samantha in the movie Her,4 the
new generation of voice-bound AI demonstrates that automated cognition has
incorporated the sociality of thinking and its affective modalities whereby
varieties in tonality, timbre, frequency, and rhythm guarantee a certain degree
of humanness. Even when Samantha decides to leave her personalized
customer, with and by means of whom she has intensified her learning about
human feelings, her voice remains trustworthy, her tone reassuring, and the
frequency of words fast enough to be soothing.
The humanness of Samantha’s voice is more akin to the entrancing call of
the Sirens,5 tempting Ulysses to abandon his all-too-human rationality, than
to the robotic sound achieved through the vocal pitching and modulation in
autotune. The latter mainly achieves the effect of a sonic human-machine
cyborg, characterised by the aural expression of the sensibility of the machine
—the aesthetic of automation—where the manipulation of small segments of
sounds reveals a certain sonic equivalence between the organic and the
inorganic. Instead, the new generation of vocal AIs has taken aural simulation
to another level. Here, the artificial voice is an expression of intelligence and
autonomous cognition, expanding beyond rather than simply remaining
equivalent to the human. Whilst the docile tone of Virtual Assistants such as
Siri, Viv, and Samantha seems to still conform to Asimov’s servo-mechanical
rules to please the master, there is something irrevocably inhuman in this
sonic synthesis of logic and calculation. If the new generation of automated
intelligences resembles the Sirens of the Mediterranean Sea, singing
inaudible frequencies that suspend Ulysses’s capacities to reason according to
moral conduct, it is because their incomprehensible speech reveals the
inhumanness of humanity, and the alienness within the human voice and
human thinking.
With automated vocality comes the realization that logical thinking,
rationality, and inferential meaning do not simply correspond to the constant
reproduction of axiomatic postulations and eternal truths. Instead, they
irrevocably confirm the realization that knowledge is incomplete and that it
involves parts of reality that are incomputable. The more perfectly the
machine is able to reproduce the human voice, the more thoroughly the
incomplete humanity of the human is revealed, beyond the comfortable
assumption of a human-machine equivalence. What is at stake here is not
simply the replacement of an optical representation of thinking—defined by
grammatical rules and syntactical connections of written words—with a sonic
regime of visceral responses. In fact, with neural networks intelligence
research, deep learning methods, and experimentation with non-deductive
logic, it is no longer possible to make this opposition between rational and
visceral knowledge. The intelligent sirens of the twenty-first century are
rather drawing out the thread of the retro-futuristic wisdom or rationality of
orality. In other words, today’s sirens are abstracting this orality from the
social complexity of speech variations. As this complexity becomes
increasingly automated, and intelligent sirens become our trustworthy
companions, one must not bemoan the end of human thinking, but wonder
about the neo-rational logic vocalizing the wisdom of knowledge.
NOTES
1. Z. Corbyn, ‘Meet Viv: the AI that wants to read your mind and run your life’, The Guardian, 31
January 2016.
2. G. Pask, Conversation Cognition and Learning (Amsterdam. Elsevier, 1975).
3. A.M. Turing, ‘Computing Machinery and Intelligence’, Mind 59 (1950), 433–60.
4. Her, dir. Spike Jonze, 2013.
5. W. Ernst, Sonic Time Machines: Explicit Sound, Sirenic Voices, and Implicit Sonicity (Chicago:
University of Chicago Press, 2016).