Another kind of AI: Alternative Intelligence

Dylan and dog in snow

We’re starting the new year with daily news feeds and timelines auguring the centrality of ChatGPT and AI to human life. The venerable New York Times publishes opinion pieces ranging from the threat to democracy posed by ChatGPT to extolling the greatest film never made . Whether opinion makers seek to make us fearful of or longing for this new ‘intelligent’ technology, consensus is squarely on its inevitability and invincibility.

Amidst this seeming consensus, headlines on the epic crash of a digital currency technology vie for only muted attention. The reasoning, best I can see, is that the crypto flame-out demonstrates a failure of greed, which is not inherent to the technology itself. Fake currency, apparently, is perfectly fine in its design. Its problem was in how it was promoted. You know, by imperfect humans.

But you don’t have to reference criminal ventures that land on the front page to know that AI won’t succeed as marketed. You witness hundreds of software and technology failures every day, if your experiences are anything like mine.

  • Every spam message that lands in your inbox is a failure.
  • Every time you have to shut down and reboot to ‘fix’ the problem with your streaming device that’s unable to move forward another frame.
  • Every time you receive a nonsensical error message on a website is a failure.
  • Every time a cashier apologizes for your wait,”Sorry, the system is just really slow today”.
  • Every time your air or train travel is disrupted (or your bags are lost) due to technical failures.

Back in the day when Microsoft was known for its standard response (“shutdown and reboot”) to fix a Windows error, exposure to software problems was mostly limited to work PCs. Over years, the software didn’t get better, but it became ubiquitous. Now you are exposed to software issues in your car, while using your home appliances, while you’re being entertained, when shopping, when traveling, using the utility infrastructure, banking, taking classes, communicating with others, and of course, working.  Or simply, while living.

Why it doesn’t get better

Technology isn’t fundamentally improving because today’s technology is based on what came before it. The latest ‘new new thing’ (a term popularized by Michael Lewis 24 years ago) doesn’t emerge tabula rasa.

Creators of new technologies act on opportunities they perceive, which are based on their experiences with other technologies that already exist. They build those experiences into the original design. And from that point forward, all of the flaws inherent to the design are baked in. Future revisions incorporate those flaws.

Which is why software doesn’t get better, and why we have learned to accept the countless failures as part of the  experience. We don’t notice so much when a checkout is awkwardly slow. We clear out the Inbox and try to improve our email controls. We pay for services to try to secure our personal data. Band-aid upon band-aid upon band-aid.

Imagining a future AI-enabled world in which the AI technology executes flawlessly is a leap of faith, since we have no experience to back it up. The technology will have multiple failure points. Its design will not be comprehensive, and its weaknesses will hobble its effectiveness. None of which means that it won’t be disruptive, simply that it won’t be reliable.

Speed of development

Fundamentally, the Achilles’ heel for technological innovation is the speed at which it’s created and adopted. Marketers are always looking for a new product to sell because bringing technology products to market at lightning speed can be immensely lucrative.

Consumers enable this rapid pace. At the end of the twentieth century, early adopters were a small sliver of the leading edge (often called the ‘bleeding edge’). Most people waited for technology to develop. Only suckers upgraded when a point release (n.0) was available, unwilling to suffer the multiple failures implicit in major releases. Such prudence is now old fashioned. Software companies have added a tax on the consumer’s time to manage updates manually and as a result, it’s commonplace to set devices to update automatically as new releases are pushed out. Adoption of new technologies is similar. Increasingly, consumers are willing to sign on to whatever is currently flaunted as hip tech. Without the time to develop, test, evaluate, and re-engineer over hundreds of thousands of iterations, and without the financial incentive to re-engineer the original design so that it’s fundamentally robust, we can only expect the fast-fashion version of technology: cheaply made, limited life cycle.

Alternative Technology that Works

Biological technology, on the other hand, evolves over millennia. Not only are incremental improvements made over time: evolution revises the design of even basic features (such as the senses) to improve their efficiency within the context of a specific animal species. As Ed Yong explains in his book An Immense World, animals have ‘endless ways of seeing’: we co-exist, but experience light and matter completely differently – so differently that perhaps for each species they are different worlds.

Individual animals exist on a spectrum of acuity for any sense, but the species itself has the technology it needs to sustain long lineages through eons of predation. It works. It’s reliable.

The accuracy in a trained dogs’ detection of specific odors (VOCs from illness, for instance) is therefore much more reliable than scientific tests, which have been developed in a relative blink of the eye. The problem with medical detection dogs isn’t the failure of the technology; it’s the failure of marketing. It’s a cheap and reliable service, which means it’s ill-suited to make a single individual as rich as Croesus. What’s the profit in that?   

Leave a Reply

Your email address will not be published. Required fields are marked *