When I sat down to write Smart Machines: How AI Is Changing Our World, I recognized the huge challenge ahead. Like most of us, I knew zip about Artificial Intelligence. How did it even work? I needed to separate truth from fiction, and then explain it in a kid-friendly style that would be fair and balanced. 

These are the top takeaways from that deep dive into the structure, function, uses, risks and benefits of AI.

Not Just One Thing.

There are many different kinds of AI.  Many of them, like machine learning (ML) software that sorts through piles of data to make predictions, can be terrific. Yes, I do want my doctors to use it to check my mammogram results.


Another type, Natural Language Processing (NLP), also has fabulous applications, e.g., translating speech into another language or into text, and back again. In 2017, a revolutionary type of AI was invented. Called transformer models, these tools became the basis for Generative AI, the shiny new software that powers chatbots and social media.

Generative AI is its Own Thing.

Generative AI significantly differs from other types of AI. Their transformer models are “black boxes” you can’t see into. No one knows how they arrive at their output. Not even their programmers.

Figure 1: The Black Box Transformer Model. Source: RPA Tech


 A Crap Shoot.

Using generative AI is risky. It’s like buying sausages from an unreliable butcher. To make the sausage, app developers use scraping tools to collect masses of data. Much of it is dirty: biased, obsolete, or just plain wrong (i.e, crap).

The slop gets fed into a backroom meat grinder—the black box. The developer stuffs it into casings (like Meta AI or SIRI). The output—some really funky sausages—gets delivered to users in a pretty package, but with a deeply rank smell. 

ChatGPT stands for:

Chat

Generative

Pretrained

Transformer

Don’t Invite These Folks to Dinner.

The hands of the folks making the generative AI sausages really are dirty. Corporations like Meta/Facebook, X, etc. knowingly produce dangerous products. They don’t follow their own ethical guidelines. They don’t obtain data fairly or protect personal data from misuse. They actively engender social unrest, anxiety and psychological dependence.

The danger is most pronounced when AI is used by younger people. Social media sites like Snapchat and TikTok are engineered to actively rewire developing brains by delivering repeated dopamine “hits.” And it’s not an accident—that’s their business model. 

(Please see a full transcript of my shocking discussion about ethics with Meta AI’s chatbot here: orcabook.com/Assets/PDFs/TeacherResource/SmartMachines_AI-Conversation.pdf)

 

The Unreliable Narrator.

Furthermore, Generative AI apps are themselves unreliable. They make stuff up, a side effect known as hallucinating or confabulating. 

They’re also easily corrupted. Hackers use techniques like Jailbreaking, Poisoning and Evasion Attacks to hijack virtually any database to deliver false information, foment fear and take control (ransomware). Scammers can also identify millions of targets for extortion, exploitation, identity theft. This means you.

 

Dr Evil on Line 1…  

What happens when governments get their grips on the AI meatgrinder? Mass surveillance is already a reality. US ICE agents, for example, are already using facial recognition AI, GPS tracking, video surveillance and more to identify individuals and pinpoint their location. They can then use that data to identify that person’s associates, friends, family members….

Environmental Costs

Despite these drawbacks, many of us continue to use generative AI tools because they’re convenient and free. But they’re not free. The environmental costs alone (data farms use huge amounts of water and electricity) are staggering.


So how do you, as an information specialist, protect your collections and your users? I’m not a Luddite, and you don’t have to be one either.  Consider these best practices from Smart Machines: How AI Is Changing Our World.

  • Educate yourself. Learn about the different types of AI and their applications so you can make informed decisions about whether using the app makes sense for you.
  • Respect copyright. Look for the Fairly Trained designation on any app you use. This ensures that the app obtains data legally and with permission from the creators.
  • Practice and model safe habits on social media. Share less information, especially personal data and photos. 
  • Review all sources for accuracy. Remember that influencers are paid shills whose goal is to get you to buy what they are flogging, and that propaganda is a profitable global business. 
  • Share your knowledge. The more people understand how AI works, the better off we all are. 
  • The jury is in: Absolutely no cell phones at school.
  • Use less tech everywhere else too. Disconnect devices from the internet wherever possible. Do you really need your fridge spying on you? None of the potential risks of a 1984-type future can happen if we simply turn everything off. 

For detailed information on sources and data used to create this article, please consult the Resources page in Smart Machines. 

Note: You can also download a free Teacher’s Guide for Smart Machines directly from Orca Book Publishers’ website. 

orcabook.com/Assets/PDFs/TeacherResource/SmartMachines_TeacherGuide.pdf