Voice Technology: Science Fiction & The Future


Voice technology has always been part of our imagination.  Back in the 40s, Dick Tracy had his watch, and voice operated computers were what made Star Trek and 2001 A Space Odyssey possible.  In large part however, voice technology remained the domain of science fiction.

Voice Technology: Science Fiction & The Future
How soon will it be before we see this ad in the classified section of our trade pubs?

In 1997, Dragon Systems released NaturallySpeaking 1.0 as their first continuous dictation product.  At that time, the error rate for speech recognition was over 43%.  Technology rapidly advanced however, and by 2001 that was down to 20%.  Speech recognition was still being treated more as science fiction than mainstream communication however, and over the next decade the error rate only declined to about 15%.  It was only when speech recognition was “disassembled” into its three major components did improvement accelerate.  There are three models that work together relative to speech recognition: the acoustic model, the pronunciation model and the language model.  The acoustic model takes the waveform of speech and chops it up in the small fragments, and figures out each sound that the person is speaking.  The pronunciation model takes the sounds and strings them together to make words, and the language model takes the words and strings them together to make sentences.  By 2016, a Microsoft research team got the error rate down to 5.9% – which matched the human error rate when a professional was hired to transcribe data.  Further advancements since then have the error rate down to about 4.9 %.  It is these most recent advances which have brought voice technology into the mainstream with products such as Amazon Alexa and Apple’s Siri.  The pace of voice technology innovation is accelerating once again as auto manufacturers and others are engaging in joint research projects to figure out how to make their products simpler and easier to use.  Once driverless technology goes mainstream, Honda envisions that their vehicles will completely change our notion of transportation.  According to one company spokesman, in the not too distant future the car is going to be an arcade, a theater, a classroom or an office – all by using your voice.  So, what does that mean for the HVAC/construction trades industry?

The trades industry have long suffered from a shortage of new talent, and that squeeze is only going to be exacerbated in coming years.  Might the advance in voice (and other) technology completely change your business and be part of the solution to this vexing problem?  Consider the following scenario in 2025.  (Six short years away)

A homeowner gets into his self driving vehicle and buckles up for the ride to work.  He grabs his tablet to check on the news, while using his voice to call his local HVAC contractor to check on the status of the technicians arrival.  Overnight, a component in his air-conditioning system had sent out an SOS because the system was no longer able to maintain the desired setpoint.  Having already queried the air-conditioning systems diagnostics panel remotely, the technician confirms her time of arrival to correct the deficiency.  Now, let’s repeat that call in 2030.  (Just 11 years away)

The air-conditioning system experiences a problem overnight which no longer allows it to maintain the homeowners desired setpoint of 0.5°F in each zone of the house.  The air-conditioning system automatically contacts its preassigned robotic technician, who queries the components and system operation diagnostics using AI technology to figure out the problem.  Within seven minutes, the problem is correctly diagnosed as a malfunctioning component, and the robotic technician scans the shops inventory only to find out the part is not in stock.  Two minutes later, the robotic technician places an online order with the local HVAC parts and supplies center to have the part delivered via drone to the homeowners residence by 9 AM the next morning.  At 9:30 AM the robotic technician arrives in a driverless vehicle and picks up the part from the secure storage unit on the homeowners porch.  The robotic technician lets itself into the house by accessing the home security system, which already knew the technician was coming.  Within 36 minutes the faulty component has been replaced, and proper operation sequence is confirmed within 10 minutes after that.  Forty-seven minutes after arriving at the homeowners residence, the robotic technician is already on its way to the next call.

Science fiction?  Perhaps, but likely only in the timeline before this becomes a reality.