OPINION: If you’ve had a play with ChatGPT, you probably have a very obvious question: why don’t digital assistants Siri, Alexa and Google Assistant have such advanced conversational skills as the chatbot that has taken the internet by storm?
After all, Amazon has had the Alexa Echo voice assistant devices on the market since 2014. Siri, Apple’s voice assistant built into its Mac computers, iPhones, smartwatches and home speakers, has been around since 2011.
Google Home devices appeared later, in 2016, but Google and Amazon spent vast sums of money subsidising the price of the hardware so they could get the devices into as many homes as possible. You can buy an Amazon Echo Dot device for about $50.
The Big Tech companies that released artificial intelligence-powered, voice-activated assistants received a trove of data yielding insights into what we want to know. Unfortunately for Apple, Amazon and Google, what we wanted to know turned out to be obvious stuff, such as the weather forecast for Auckland, or screening times at the local movie theatre. Amazon, to its dismay, discovered that people didn’t use Echo to search for and buy things in its massive online store.
I use the Google Home devices around my house to stream music and control smart LED lights in each room. These are handy features, but after a decade of solid investment, you would expect these gadgets to have evolved into something more useful.
As Microsoft boss Satya Nadella rather smugly put it last month, these voice assistants are “dumb as a rock”. At least Microsoft had the sense to give up on its own one, Cortana, seeing the inability of its rivals to make voice assistants useful – or profitable. Microsoft’s integration of OpenAI’s ChatGPT-like features into its Bing search engine and suite of Office products looks set to be far more useful than anything the voice assistant revolution has produced.
That’s because they are based on different types of AI. The traditional voice assistants understand a limited, if large, list of questions and requests. They draw on a wide range of internet-based sources in order to try to answer your question, but too often will come up with the unsatisfactory answer, “Sorry, I don’t have any information about that”.
ChatGPT instead is generative AI, powered by large language models. It is trained to recognise and generate answers based on enormous sets of data harvested from the web. The result is a much more conversational and, dare I say it, human response. The downside is that it will make stuff up just to keep you happy, at least until OpenAI’s content censor steps in. Its answers need thorough fact-checking, but overall, it is incredible at producing sensible, interesting and even humorous responses.
Google, in particular, has been working on generative AI for years. So why didn’t it build it into Google Assistant and its Home devices? That’s an answer disgruntled Google shareholders are now asking. Ultimately, the tech companies thought we’d use those devices in a different way, to organise our lives, automate routine tasks and make purchases.
The real value, it turns out, is having a chatbot that can be a good conversationalist. We can expect conversational AI to be hastily built into digital assistants, so don’t junk your Echo or Google Home device just yet. A conversational AI mode for my Google Home gadgets would involve just a software upgrade.
But ChatGPT is currently text-based. Running it as a voice assistant introduces more scope for errors and misunderstandings, particularly when the Kiwi accent is involved.
Deep cuts at Amazon in recent months have taken aim at its loss-making Alexa voice assistant division. But there’s no way it will give up on voice assistants. The success of ChatGPT sends a message to Amazon and its rivals that we are ready to take the next step on the AI journey. But they will be kicking themselves that they missed the opportunity to lead the way.
