The buzz used to be about what it would be like when a device could do something that humans do. It might be assembly line work or mowing the lawn, self-driving cars or self-checkout lines, robotic bartenders or robotic surgeons. The thought was that tasks which were repetitive and precise and physical could be better handled by machines.
Then reality set in and it went from exciting or worrying to merely interesting. Sure, it's cool to watch giant machines spin and twist as they assemble cars, or remote-controlled forklifts race through a warehouse and pluck pallets from shelves. But all are overseen by flesh and blood managers, and computerized devices came to be seen more as high-tech tools and not replacements.
Still, it freed us people to have more time to do what machines couldn't, which was to think and create. And while the ability of computers to ever be able to do the first is up for debate, the second seems to be in play. Story after story has touted how computer-generated responses to queries now can come back not as a list of web sites, but as essays and articles that could plausibly have been written by a person. And so the "will be replaced by robots" cautionary tale of the month is not about humanoid receptionists or robotic security guards, but about artificial intelligence, or A.I.
In and of itself, A.I. is not new. The first use was in 1942, when Alan Turing created a machine to crack the German Enigma code. Breakthroughs continued: the first chatbot appeared in 1964 as MIT's Eliza, followed by ALICE (Artificial Linguistic Internet Computer Entity) in 1995 and IBM's DeepBlue in 1997. That led to Watson, Siri and Alexa as well as your Roomba vacuum cleaner, all of which were cause more of amazement as opposed to alarm. But with the rollout of OpenAI's ChapGPT and Google's Bard, systems which seemingly appear to be able to give human-level cognitive responses without the benefit of opposable thumbs, all of our hair is suddenly on fire.
To be very clear and as stated again and again, these are super sophisticated pattern generators and not independent thinkers. Trained on millions and millions of examples scraped from our world, they put one logical word after another to build up a response. The result mimics human writing and speech, but they are just that: mimics. That's why one of the biggest problems occurs when they are asked to cite their work. The names for web sites have no real logic, and so an A.I. response to your question will give its reference as htttps://web/asksfd723mafp when no such thing exists. Those are called hallucinations, which imply extrapolations from a real base. They could just as easily be called what they are: gibberish.
The question is not so much if A.I. can mimic a human response but rather can it make a human choice. Every day we make countless decisions about our actions: whether to get out of bed when the alarm rings, which route to take to work, which emails to respond to and which to ignore. You've made a several in the past 60 seconds, merely by deciding to read this piece and even to keep going this far. In fact, scientists at Cornell say we make 221 decisions each day just about food. And while you may enlist a computer's guidance for any or all of those choices, odds are you won't be ceding final authority on those to a machine.
And even if you did? Research commissioned by the psychology-based app Noom says that you won't buy it anyways, that the choices we make are hardly carved in stone. We second (and even third and fourth) guess what we select. The study found that the average adult admits to changing their mind twice per decision, with more than 11% doing so five or more times. That goes for what to watch on TV, what to wear and what to buy (or not).
It would seem that the key for A.I. to appear human is not so much to be intelligent as to be indecisive. That's why the tell as to whether "to be or not to be" was written by Siri or Shakespeare might turn out to be not that Hamlet couldn't pose the question, but that he couldn't make up his mind.
-END-
Marc Wollin of Bedford changed his mind more than a few times writing this. His column appears regularly in The Record-Review, The Scarsdale Inquirer and online at http://www.glancingaskance.blogspot.com/, as well as via Facebook, LinkedIn and Twitter.
No comments:
Post a Comment