Qwertyman No. 21: AI in the House

Qwertyman for Monday, December 26, 2022

I’D BEEN wanting to write about this for a long time, since last year when my artist-friends first alerted me to the amazing new possibilities being opened by artificial intelligence (AI) in such traditional fields as painting. Before that, like many people, I’d thought of AI in terms of subjects like warfare, medicine, and gaming. It had to be only a matter of time before the technology was ported over to the arts—not just to painting, but to creative writing and music, among other pursuits.

If you don’t know how AI works, just think of it this way (which is the way a lot of people, many of them not even artists, are having fun these days). You download a software program called an “AI generator” (such as starryai for painting, Rytr for creative writing, and AIVA for music), then put it to work by demanding that it produce “a portrait of Jose Rizal in the style of Van Gogh.” Minutes later (or just seconds if you pay), you’ll get what you asked for. What your computer (or in fact, many other computers working together) just did was to run a search for all the images of “Jose Rizal” it could find, then establish what “the style of Van Gogh” means in terms of brush strokes, colors, and so on, and apply one to the other. It’s all about “algorithms” or instruction sets that get sent out and executed until the desired outcome pops up.

That sounds like harmless entertainment, and much of AI is or appears to be, but it isn’t hard to see how even a touch of mischief (not to mention loads of malice) can quickly change the complexion of things. Driven by a fear of computing or what we might call Faustian science, popular culture is replete with examples of AI gone berserk, from HAL 9000 in 2001: A Space Odyssey to the master kidnapper Sark in Tron. But positive applications such as those employed in medicine, where AI helps analyze CT scans, MRIs, and X-rays with a sharper eye than a human radiologist sees with, continue to push AI research forward. IBM itself declares that “At this point, there is little doubt that AI will become a core part of the digital health systems that shape and support modern medicine.”

My own first brush with AI happened last week when I stumbled on an article in the New York Times titled “A New Chat Bot Is a ‘Code Red’ for Google’s Search Business.” Intrigued, I read on and found out about ChatGPT, one of many new programs that are threatening to replace Google’s famous search engine. We all know how Google works: you enter a search term, you get a long list of links to possible answers. Chat bots cut through that mess and give you a short, clear, grammatically correct answer (although nobody can guarantee that the answer itself is correct). That’s good if you want a quick reply to a question, but I can just imagine the epidemic of college term papers cutting and pasting unsourced chat-bot answers with gay abandon. 

All this led me to wonder how chat bots would respond to truly important questions, so I decided to download ChatGPT to my iPhone, and to ask it some test questions that would tell me about itself as much as what I was asking about. Below are my questions and the answers that artificial intelligence produced, verbatim (I purposely asked the first question to see if AI was capable of some kind of self-awareness):

1. Q. What are the dangers of artificial intelligence? A. 1. Unintended Consequences – Algorithms designed to reach specific goals could have unintended consequences that produce unforeseen results. 2. Job Loss – AI could displace workers by automating jobs, leading to severe economic disruption in the near future if nobody is trained to take over for the machines. 3. Black Box AI – AI systems are often difficult to understand, making it hard to debug the system when something goes wrong. 4. Over-Reliance on AI – Dependency on AI could lead to cognitive laziness, where humans rely too heavily on algorithms for decision-making and problem-solving. 5. AI Biases – AI systems trained on biased data and algorithms can end up replicating the same flawed assumptions and stereotypes. 6. Security Breaches – Malicious actors could exploit vulnerabilities in AI systems to commit cyber-attacks, data breaches and fraud.

2. Q. What’s wrong with the Philippines? A. The Philippines faces a wide range of problems that must be addressed in order to progress forward. These issues include corruption, overpopulation, environmental destruction, poverty, inequality, and unequal access to education. Additionally, Filipino infrastructure and public services are in dire need of improvement in order to create a more livable and equitable society. Other issues such as illegal drugs, crime, and terrorism also remain major problems facing the Philippines.

3. Q. When should a country have a sovereign wealth fund? A. A country should have a sovereign wealth fund when it has a large amount of assets (such as revenues from oil exports) that can be used for investments and for stabilizing government finances during economic downturns.

4. A. How can we reduce corruption in government? A. The best way to reduce corruption in government is to increase transparency and accountability. This could include strengthening laws and regulations to ensure that government officials are held to high standards, creating independent oversight and auditing systems, and engaging civil society and community members in governance processes. Additionally, investing in training for government officials on ethics, anti-corruption policies and procedures, and proper financial management can help to reduce corruption. Finally, engaging the public to report and discuss any allegations of corruption can help to both curb and address existing instances of corruption.

5. Q. Can volcanoes be punctured? A. No, volcanoes cannot be punctured. Volcanoes are made of molten rock, known as magma, that is too hot and too deep to be pierced with any known tool.

Now, you tell me if those answers sound any worse than the malarkey we’ve been hearing from some of our esteemed senators and congressmen, who cost billions to maintain, feel unbound by the rules of logic and grammar, and can take forever to give you a straight answer. Of course, they can well afford to buy a program like ChatGPT in aid of legislation (it’s free for three days, then P499/mo.). But then, why resort to chat bots when there’s already so much artificial intelligence going around in both august chambers?