Artificial intelligence

By Don Smolenski, Contributing Editor | TLT Machinery June 2024

Is this the new frontier?

Artificial intelligence (AI) research in the 1950s explored topics like problem solving and symbolic methods. In the 1960s, the U.S. Department of Defense started training computers to mimic basic human reasoning.1,2 The Defense Advanced Research Projects Agency (DARPA) produced intelligent personal assistants in 2003, decades before Siri or Alexa. I hear or read something about AI virtually every day. The term artificial intelligence was coined in 1956, and AI has become more popular today by virtue of increased data volumes, advanced algorithms and improvements in storage and computing power. It seemed to me that it would be a good idea to obtain at least a rudimentary understanding of what it is all about. AI can be defined as the ability of a machine or software to perform the cognitive functions (thinking!) that we usually attribute to human minds. In truth, humans have utilized machines (i.e., the wheel, the lever) for centuries, greatly enhancing our productivity. 

AI is basically a subsection of computer science. Specifically, it deals with studying machine learning and developing intelligent machines. In AI, machines can learn from experience and make adjustments to new inputs. AI often relies on deep learning, which teaches computers to process data much like the human mind does. It also may employ natural language processing, which employs text analytics that can be integrated into existing data and support several languages. Computers can thus be trained to accomplish very specific tasks using large databases and recognizing data patterns. This basic work enables automation and formal reasoning in computers today, including decision support and smart search systems that can segment and complement human abilities. 

So, what are the downsides to AI? Well, first, AI is only as good as the data it utilizes (the old adage of “garbage in, garbage out” still very much applies). Could I use it to write my columns? Potentially, yes, but what I would give up is a true understanding of the things that I research. Plus, TLT now has an AI policy. What are the dangers of AI? Could computerized robots take over the world, for instance? Probably not. Another real and very important concern, however, is the issue of ethics, and how they would be addressed. 

According to the current system of classification, there are four primary AI types:3 
• Reactive. Task specific—an input always delivers the same output, because they have no memory. Machine learning models are often reactive. 
• Limited memory. These machines can review the past and track specific situations or objects incorporating these data into the AI models so that it uses past and present data. 
• Theory of mind. Having the ability to infer knowledge and intentions of others, and use this information to predict actions and behaviors. 
• Self-aware.4 These artificial systems have self-perception beyond rule-based decisions and pattern recognition; these AI systems can comprehend their own intentions and actions. 

So, what’s most important going forward? “It really is who has the data. That’s who will be the king,” according to Harper Reid, Technology Pioneer.5

REFERENCES
1. “The AI Chronicles” Podcast (July 2023), “History of Artificial Intelligence.” Available here.
2. Click here.
3. 4 Types of AI: Getting to Know Artificial Intelligence, click here.
4. Tyagi, A. (Oct. 24, 2023), “What if AI becomes self-aware?” Available here.
5. Click here.
 
Don Smolenski is president of his own consultancy, Strategic Management of Oil, LLC, in St. Clair Shores, Mich. You can reach him at donald.smolenski@gmail.com.