But Chatbot Algorithms Still Aren’t Substitutes for Human Imagination
The great virtual assistant race has been underway for more than a decade and the emerging winner isn’t Siri, Alexa or the nameless voice that provides directions when you’re driving. The run-away winners are chatbots, which still have a long way to go to catch humans. More on that later.
Virtual assistants and chatbots both rely on artificial intelligence. What separates them is the design of chatbots that lets them advance at lightning speed, contrasted with cumbersome designs for virtual assistants that can require days or weeks to update. In today’s ‘digiverse’, days and weeks are an eternity.
The upshot is that chatbots have become creative assistants, leaving Siri and her pals in the dust providing weather updates, setting alarms and cuing up favorite music titles. Not surprising then that a significant percent of Amazon’s 18,000 layoffs last year who staffers who worked on Alexa. Of Google’s 12,000 layoffs, the team working on virtual assistants lost 16 percent of its engineers.
The emergence of ChatGPT has sent shock waves through the world of high tech. While many worry about its potential for plagiarism, big tech companies recognize it can be a severe blow to their virtual assistant businesses and are scrambling to regroup in AI.
Some AI futurists foresee chatbots and virtual assistants playing complementary roles. Users can speak to their virtual assistants,. which, in turn, can tap into the richer realms accessible through chatbots. It’s like asking the waiter to tell the chef to make a five-course meal.
Perhaps more likely, Apple, Amazon, Google and Microsoft will shift to their own chatbots. Microsoft has invested in OpenAI, ChatGPT’s developer, with the goal of incorporating chatbot capability into its Bing search engine. Google is testing a chatbot called Bard.
The Key Differences
Chatbots rely on “large language models” trained to pore through enormous online data sets and construct complete sentences. Virtual assistants use “command-and-control” models that can recognize and respond to a finite list of questions or requests.
Chatbots can be expanded by constantly feeding them more data sets. When introduced, ChatGPT stumbled when asked to write a haiku, a three-line, 17-syllable unrhymed poem. A few months later, it quickly churned out haikus.
Siri’s code took weeks to update with basic features. Adding new phrases required rebuilding the entire database. Amazon and Google bet on uses for their virtual assistants that didn’t pan out commercially. A Microsoft official recently described voice assistants as “dumb as a rock”.
Chatbots Aren’t Perfect
Improvements in chatbots are exponential but that doesn’t mean they are flawless or their use is pristine. Chatbots are becoming more accurate, yet their algorithms still make things up. Worse, chatbots may lack the capability to distinguish truth from lies or original from fakes.
Chatbots are highly evolved virtual assistants, not human colleagues. They can fetch information, rough out narratives and save a lot of legwork. They can’t replicate a human perspective or emotion.
Giddiness over what chatbots can do has fueled a funding frenzy for startups, which is good for entrepreneurs, but not necessarily content creators. What the world needs now is not rote information, but discriminating perspective. Artificial intelligence may be able to deliver that someday, but not yet and maybe not for a long time.
AI may be ingenuous but lacks ingenuity.
The Human Voice
With ossifying polarization, the last thing we need are tools that can mass produce disinformation. AI is a tool, and like all tools it can do good or do mischief, much like fast cars, chainsaws and laptop computers. The difference between delight and disaster lies in the hands and heads of its users.
While chatbots can find, organize and put oat our fingertips an incredible depth of information, its work product still lacks the human judgment that comes from lived experience, value systems and personal insight. As humans, we aren’t fully cognizant how our brains process information, detect relevant patterns and form judgments. It would be hard to farm that out to AI algorithms.
In his book Until the End of Time, physicist Brian Greene devotes a chapter to the origin of human language, which he links to evolutionary adaptation:
“With language we can articulate a vision….We can conjure imagery, authentic and fanciful, We can pass on hard-earned knowledge, substituting the ease of instruction for the difficulty of discovery. We can share plans and align intentions, facilitating coordinated action. We can combine individual creative capacities into an immensely consequential communal force. We can look into ourselves and recognize that, though shaped by evolution, we are able to soar beyond the needs of survival. And we can marvel at how a carefully arranged collection of grunts, glides, fricatives and stops can convey insight into the nature of space and time or provide an affecting portrait of love and death.”
Greene says no one can conclusively prove who invented language or how it evolved. What we do know is language has been critical to human accomplishment and survival. That’s not a role that can be offloaded to machines, no matter how intelligent we make them.
Artificial intelligence may summon vast amounts of existing knowledge, observations and art and produce facsimiles of creativity, but it can’t replicate experience, imagination or human perception. AI may be ingenuous but lacks ingenuity. Machines cannot “feel a thought” or experience an emotion that yields enlightenment, perspective or empathy. AI has an algorithm not an imagination.