So-called.

Media Objects, a new podcast produced by The World According to Sound and Cornell Media Studies scholars, adds this skeptical tag before the phrase “artificial intelligence” and “generative artificial intelligence” so many times in two episodes on the topic it’s started to stick in my mind as I read or listen to articles from other sources. 

This insightful and deeply researched podcast has encouraged me to do something I see few doing: think critically about AI.

Not doubt. Not pretend it doesn’t exist. Not denounce or ignore. Just THINK.

It’s something many, many users of AI are not doing – choosing instead to gulp down a narrative fed to them by the companies cashing in on new packaging of technology that’s not actually that new.

And it’s something the artificial intelligence [so-called] itself cannot do.

Think. Reason. Comprehend. Communicate. CREATE.

I appreciate Media Objects episodes 5: Artificial Intelligence, The Metaphor and 6: Artificial Intelligence, The Reality, for the way they pull back the curtain, putting AI into historical context, revealing the intellectual property that’s gone, without permission, into major AI products, reaffirming the technological truths feeding into their products, and re-centering human intelligence as the wonder it is.

I highly recommend listening. You can catch both parts on Bay Area radio KALW’s Bay Made show this week, starting August 4 at 11:30 a.m. PDT/1:30 p.m. eastern; on the Media Objects website, or wherever you get your podcasts.

Some highlights for me:  

• The popular myth of artificial intelligence tells us that AI thinks like a human, only better, faster, on a grander scale. However, AI’s capacity for language, through the large language model (LLM), is fundamentally different than human. It’s based on prediction, rather than the complex of processes that go into human thought and communication. “Language is not numbers. Communication is not computation. Thought can’t be reduced to probability,” the podcast points out. “But that’s exactly what it is to an LLM.”
• The myth that computers can be intelligent in the way humans are leads to a paradox: “The idea that a computer could do and think like a human, yet also operate and calculate like a machine. That somehow our creative, associative, imprecise, emotive, messy and often illogical intelligence, that we are not close to quantifying or explaining, will be replicated by computation, by probability.” This contradiction becomes clear if you stop and think for a moment.
• Humans assign human characteristics to objects around us. As the mom of a 4 year old, I’m immersed daily in the worlds of Pixar movies Toy Story, Cars, Finding Nemo, A Bug’s Life, and it’s true—we love to make inanimate objects walk, speak and interact. But this lovely attribute that allows us to make up vivid stories, enjoy childhood and get attached to our automobiles and office plants also fools us into thinking a machine that can talk, walk and interact like us IS like us. It’s not.
• Neither is so-called artificial intelligence as new and revolutionary as the marketing myth would have us believe. As the podcast points out, LLMs have been integrated for years into search engines, photo processing, autopilot systems, weather simulators, social media algorithms and Youtube recommendations. “Long before you ever heard of so-called generative AI, your life was already filled with content altered, generated or served up to you by LLMs and other machine learning models. We just weren’t being marketed so aggressively yet with the term ‘artificial intelligence.’”

As computer scientist Pedro Domingos said, “People worry that computers will get too smart and take over the world. But the real problem is that they’re too stupid, and they’ve already taken over the world.”

Yet so-called artificial intelligence – a misleading metaphor for machine processes – is being sold to us by companies that stand to make fortunes. Already are.

The second episode, Artificial Intelligence, The Reality, focuses on how so-called generative AI is not revolutionary, but rather a logical next step in trends that have been going on for decades, even centuries: automation of labor, growth of mass media, increases in monopoly power. It also takes a clear-eyed look at the intellectual property it’s consumed to create its models; and the energy, water, and mineral resources it takes to support its physical infrastructure.

“By understanding this context, we can get a much clearer picture of what so-called generative AI actually is, what the companies behind it are really up to, and all the ways it can affect our lives,” the episode says.

AI is at work in our world – not all hype and not all for the bad. Last week, I wrote an announcement of the National Science Foundation’s AI Materials Institute (NSF AI-MI), led by Cornell, through which top materials researchers will use AI (and partner with researchers studying AI – Lord knows we need more of that) to generate materials for sustainable energy, advanced electronics, and quantum technologies. This is a positive advance, harnessing AI to intentionally create (as opposed to waiting to discover) new materials. Some of the world’s most dedicated researchers are using AI in a specific, focused context, to work through the reems and reems of data produced as human researchers try to understand the nature of things like electron behavior. One researcher told me AI will help his lab speed up tasks 10 to 100 times that would take very smart humans a year.

We don’t have to swallow the hype – positive or negative. I prefer a thoughtful approach to anything, especially something that will supposedly turn my life upside down. Media Objects, which is rooted in the humanities, provides this perspective on this thing that’s called – but actually isn’t – intelligent.