noelward
Well-known member
Is AI as Smart as Your Dog?
By Noel Ward, Editor@Large and AI Skeptic
Dogs are said to possess the intelligence of the average 2-to-5 year-old human. Our Corgi seemed to be at the high end of that range while our Springer was definitely at the lower end. Come to think of it, I once had a boss with about the same smarts as the Springer, except he could talk. Unfortunately.
OK, the dogs did things dogs are good at. AI is not so different. It only appears smart because it is extremely fast and may show people things they have not thought of. This does not make it smart. The illusion of intelligence is that AI cites stuff people probably don’t know. They have not—in AI terms— been “trained” on a vast amount of info combined with immensely complex connective algorithms.
Is there a thought process?
Some dogs (certain herding breeds are good at this) can retrieve an object you identify by name if they have seen it before and know what it is called. Many dogs seem to think, at least what I consider reactively. This sounds a bit like AI on a furrier, friendlier, smaller and probably less complex scale. As reader “MailGuru” recently noted, “It sounds to me like it [AI] is still just a very large data storage and retrieval system. No actual thought process is involved. It simply stored that information and regurgitates it on demand.” Huh.
Being skeptical, I wonder if AI can be sure of anything or is it simply reactive? Evar curious, I asked AI about the technologies and physics involved in a scene in a novel I’m writing. AI was mostly reactive yet in all fairness, I learned some things that would have taken hours to find and validate. I fact-checked what AI claimed, including the technology, physics and other elements. Turns out AI was not bad at all. It was fast and convenient, a useful tool. On the other hand, it was not unlike going from a hammer to a pneumatic nail gun. Or from a hand drill to a battery-powered one. Or using a computer instead of writing longhand on paper with a pen. Faster, more efficient and saves a lot of labor/time but is not magic. Nor is AI: It’s just a computer program. And not intelligent.
So what is intelligence?
You already know AI is trained on vast amounts of topical and technical information using LLMs (Large Language Models). This constitutes far more info than a human could find, read, absorb, and categorize in a lifetime. More critically, the training includes complex algorithms that mix Linear Regression, Calculus, Probability/Statistics, and something called Discrete Math, all herded by computer programming languages such as Python. The algorithms pull fragments of information together to offer up a response that is accompanied by disclaimers about accuracy.
And there’s a problem with that. “Large language models are not reasoning machines. They’re plausibility engines. It’s not just that they don’t test their outputs to make sure they’re correct or logical, or that they fail to do so in certain instances. They can’t. And they’ll never be able to on their own,” noted Princton’s Zeynep Tufekci in a June 30 New York Times column.
Testing responses is the job of the algorithms. But I took a lot of statistics, worked with it in the real world, and know the responses being tested are probabilities, all based on AI’s training. Information, especially when guessed at, is not intelligence. Moreover, AI really doesn’t know what it’s “talking” about. I queried Gemini, Google’s AI flavor which pointed out that it lacks the practical experience of actually doing something —admitting this to be a weakness— despite arriving at one’s computer or phone as an unbiased response. A problem here is that practical experience is a huge part of learning, knowledge, and describing possible results.
On the other hand, AI can also be very helpful in some specialized areas of science and medicine because the algorithms can pull fragments of information together in ways humans cannot. First, people don’t have access to all the info and second, are generally unable to correlate disparate facts effectively. Our brains simply aren’t wired that way. AI is great at this and it sure beats days or weeks of research.
So envision this
Think of AI a perpetually growing 3-dimensional spreadsheet with billions of rows and columns of cells containing the entirety of human knowledge. Algorithms tie these cells together to provide answers that may be correct, pulling together related and semi-related info from many sources, selecting what may be more and less relevant, then handing you a version of what it has collected, and making suggestions for a deeper dive. But probe correctly and AI will tell you it has no clue whether its info is right or wrong, one reason AI responses come with a disclaimer. Once it lays out what it finds, AI may ask if you would like its responses configured as a business plan, presentation, proposal or whatever. The depth of info provided and the speed in which the response is compiled can make this seem miraculous. But fast is not the same as accurate. Do you prefer an employee provide a fast answer right now, or an accurate one tomorrow morning?
Jobs anyone?
In its June 9th issue the New York Times ran a story about what AI might mean for jobs. Targeting NYT readership, the article dealt primarily with “knowledge workers” who in one way or another includes relatively recent college grads. As I read the story, I started thinking how lots of smart and skilled people don’t go to college or attend briefly before deciding it’s not for them. Others with college diplomas wind up in fields for which they have no degree and learn on the job, some rising to management roles. The thing is, humans are generally smart and most people have several things they do well in addition to whatever job(s) they hold. Three good friends come to mind, all of whom happen to be exceptionally skilled woodworkers. The geophysicist builds cabinets. A veterinarian restores wooden boats. The lobsterman is an all-around carpenter and generally handy. All learned wood-smithing by doing. Meanwhile, I I got through college being a pro-grade auto mechanic and once made a living delivering yachts and captaining sailing charters. Most of the knowledge we use in our lives did not come from college. And it won’t come from AI.
This provokes some questions: What, if anything, does AI replace? In what ways? At what point in AI’s intrusion does human judgement come into play? How can AI be combined with practical knowledge acquired last week applied to so you can make relevant decisions or choices? What jobs might it displace? And so on.
Practical knowledge
Let’s assume you’re using AI for one or more elements of your printing business. You already know a lot about your specific operation and that AI’s knowledge is only generic. AI for instance, doesn’t necessarily know much about UV or varnish coatings, just that they can be applied. Your pressman knows far more. You probably do too.
Instead of relying on AI, find out what your employees know about your business and market that you don’t. You may be surprised. Talk with customers too. And other printers. I work with an association of privately held transactional and direct mail printers and their vendors. Together, they account for about nine percent of all US mail. Although they routinely compete, they also share information, mixing practical operational experience and human intelligence, no computer required. Most also use AI to accumulate and track info, aid in planning, help respond to RFPs, and more. Yet they tell me the most valuable info comes from the shared thinking, experience and advice of peers who’ve faced similar challenges. The vendors say they learn from group members all the time. Such on-the-street knowledge is something AI doesn’t possess. At least not yet.
Not actual intelligence
AI is not someone you can have a beer with or kick around ideas over a barbeque. Use it to find out things you know little about but fact check what it claims and balance that info with practical advice from colleagues. Along the way, be aware that some “knowledge” may come from savvy AI users. One AI-advocating colleague found AI can frame its responses based on user queries, even though the info submitted by the user may be blatantly wrong. This means an incorrect assumption on your part can lead to errors in what AI tells you. In my colleague’s case (he was experimenting) he found his deliberately erroneous info could influence AI’s responses and point a user in the wrong direction. Of course, much the same can be done using data from Excel or QuickBooks. Remember, AI is just a computer program responding to input. It’s a machine with reactive—not reasoning—capabilities. Maybe AI should stand for “Artificial Information.” Just kidding. Mostly.
You’ll have plenty of your own thoughts, questions and answers, so this could be an interesting discussion instead of a mere web article. I know some readers will feed this diatribe into an AI engine. Have a fine time, and as tourists say in New Orleans’s French Quarter, laissez le bon temps rouler. Skeptical as I am, I use AI to find out things I don’t know. Then I check it against real-world knowledge and experience. It’s useful and entertaining all at once! Sort of like my late Corgi.
Anyway…
My skepticism aside, I encourage you to use AI in your business and share your experiences. Just be smart about using it. It may show you things you have not thought of or new ways of using what you already know. As AI becomes more refined you may use it in ways you have not yet envisioned, hopefully saving money, being more profitable, more competitive, or operating more efficiently. Like it or not, AI is a good tool that when mixed with human intelligence is becoming part of business as usual. Or will be once it is refined a bit more.
By Noel Ward, Editor@Large and AI Skeptic
Dogs are said to possess the intelligence of the average 2-to-5 year-old human. Our Corgi seemed to be at the high end of that range while our Springer was definitely at the lower end. Come to think of it, I once had a boss with about the same smarts as the Springer, except he could talk. Unfortunately.
OK, the dogs did things dogs are good at. AI is not so different. It only appears smart because it is extremely fast and may show people things they have not thought of. This does not make it smart. The illusion of intelligence is that AI cites stuff people probably don’t know. They have not—in AI terms— been “trained” on a vast amount of info combined with immensely complex connective algorithms.
Is there a thought process?
Some dogs (certain herding breeds are good at this) can retrieve an object you identify by name if they have seen it before and know what it is called. Many dogs seem to think, at least what I consider reactively. This sounds a bit like AI on a furrier, friendlier, smaller and probably less complex scale. As reader “MailGuru” recently noted, “It sounds to me like it [AI] is still just a very large data storage and retrieval system. No actual thought process is involved. It simply stored that information and regurgitates it on demand.” Huh.
Being skeptical, I wonder if AI can be sure of anything or is it simply reactive? Evar curious, I asked AI about the technologies and physics involved in a scene in a novel I’m writing. AI was mostly reactive yet in all fairness, I learned some things that would have taken hours to find and validate. I fact-checked what AI claimed, including the technology, physics and other elements. Turns out AI was not bad at all. It was fast and convenient, a useful tool. On the other hand, it was not unlike going from a hammer to a pneumatic nail gun. Or from a hand drill to a battery-powered one. Or using a computer instead of writing longhand on paper with a pen. Faster, more efficient and saves a lot of labor/time but is not magic. Nor is AI: It’s just a computer program. And not intelligent.
So what is intelligence?
You already know AI is trained on vast amounts of topical and technical information using LLMs (Large Language Models). This constitutes far more info than a human could find, read, absorb, and categorize in a lifetime. More critically, the training includes complex algorithms that mix Linear Regression, Calculus, Probability/Statistics, and something called Discrete Math, all herded by computer programming languages such as Python. The algorithms pull fragments of information together to offer up a response that is accompanied by disclaimers about accuracy.
And there’s a problem with that. “Large language models are not reasoning machines. They’re plausibility engines. It’s not just that they don’t test their outputs to make sure they’re correct or logical, or that they fail to do so in certain instances. They can’t. And they’ll never be able to on their own,” noted Princton’s Zeynep Tufekci in a June 30 New York Times column.
Testing responses is the job of the algorithms. But I took a lot of statistics, worked with it in the real world, and know the responses being tested are probabilities, all based on AI’s training. Information, especially when guessed at, is not intelligence. Moreover, AI really doesn’t know what it’s “talking” about. I queried Gemini, Google’s AI flavor which pointed out that it lacks the practical experience of actually doing something —admitting this to be a weakness— despite arriving at one’s computer or phone as an unbiased response. A problem here is that practical experience is a huge part of learning, knowledge, and describing possible results.
On the other hand, AI can also be very helpful in some specialized areas of science and medicine because the algorithms can pull fragments of information together in ways humans cannot. First, people don’t have access to all the info and second, are generally unable to correlate disparate facts effectively. Our brains simply aren’t wired that way. AI is great at this and it sure beats days or weeks of research.
So envision this
Think of AI a perpetually growing 3-dimensional spreadsheet with billions of rows and columns of cells containing the entirety of human knowledge. Algorithms tie these cells together to provide answers that may be correct, pulling together related and semi-related info from many sources, selecting what may be more and less relevant, then handing you a version of what it has collected, and making suggestions for a deeper dive. But probe correctly and AI will tell you it has no clue whether its info is right or wrong, one reason AI responses come with a disclaimer. Once it lays out what it finds, AI may ask if you would like its responses configured as a business plan, presentation, proposal or whatever. The depth of info provided and the speed in which the response is compiled can make this seem miraculous. But fast is not the same as accurate. Do you prefer an employee provide a fast answer right now, or an accurate one tomorrow morning?
Jobs anyone?
In its June 9th issue the New York Times ran a story about what AI might mean for jobs. Targeting NYT readership, the article dealt primarily with “knowledge workers” who in one way or another includes relatively recent college grads. As I read the story, I started thinking how lots of smart and skilled people don’t go to college or attend briefly before deciding it’s not for them. Others with college diplomas wind up in fields for which they have no degree and learn on the job, some rising to management roles. The thing is, humans are generally smart and most people have several things they do well in addition to whatever job(s) they hold. Three good friends come to mind, all of whom happen to be exceptionally skilled woodworkers. The geophysicist builds cabinets. A veterinarian restores wooden boats. The lobsterman is an all-around carpenter and generally handy. All learned wood-smithing by doing. Meanwhile, I I got through college being a pro-grade auto mechanic and once made a living delivering yachts and captaining sailing charters. Most of the knowledge we use in our lives did not come from college. And it won’t come from AI.
This provokes some questions: What, if anything, does AI replace? In what ways? At what point in AI’s intrusion does human judgement come into play? How can AI be combined with practical knowledge acquired last week applied to so you can make relevant decisions or choices? What jobs might it displace? And so on.
Practical knowledge
Let’s assume you’re using AI for one or more elements of your printing business. You already know a lot about your specific operation and that AI’s knowledge is only generic. AI for instance, doesn’t necessarily know much about UV or varnish coatings, just that they can be applied. Your pressman knows far more. You probably do too.
Instead of relying on AI, find out what your employees know about your business and market that you don’t. You may be surprised. Talk with customers too. And other printers. I work with an association of privately held transactional and direct mail printers and their vendors. Together, they account for about nine percent of all US mail. Although they routinely compete, they also share information, mixing practical operational experience and human intelligence, no computer required. Most also use AI to accumulate and track info, aid in planning, help respond to RFPs, and more. Yet they tell me the most valuable info comes from the shared thinking, experience and advice of peers who’ve faced similar challenges. The vendors say they learn from group members all the time. Such on-the-street knowledge is something AI doesn’t possess. At least not yet.
Not actual intelligence
AI is not someone you can have a beer with or kick around ideas over a barbeque. Use it to find out things you know little about but fact check what it claims and balance that info with practical advice from colleagues. Along the way, be aware that some “knowledge” may come from savvy AI users. One AI-advocating colleague found AI can frame its responses based on user queries, even though the info submitted by the user may be blatantly wrong. This means an incorrect assumption on your part can lead to errors in what AI tells you. In my colleague’s case (he was experimenting) he found his deliberately erroneous info could influence AI’s responses and point a user in the wrong direction. Of course, much the same can be done using data from Excel or QuickBooks. Remember, AI is just a computer program responding to input. It’s a machine with reactive—not reasoning—capabilities. Maybe AI should stand for “Artificial Information.” Just kidding. Mostly.
You’ll have plenty of your own thoughts, questions and answers, so this could be an interesting discussion instead of a mere web article. I know some readers will feed this diatribe into an AI engine. Have a fine time, and as tourists say in New Orleans’s French Quarter, laissez le bon temps rouler. Skeptical as I am, I use AI to find out things I don’t know. Then I check it against real-world knowledge and experience. It’s useful and entertaining all at once! Sort of like my late Corgi.
Anyway…
My skepticism aside, I encourage you to use AI in your business and share your experiences. Just be smart about using it. It may show you things you have not thought of or new ways of using what you already know. As AI becomes more refined you may use it in ways you have not yet envisioned, hopefully saving money, being more profitable, more competitive, or operating more efficiently. Like it or not, AI is a good tool that when mixed with human intelligence is becoming part of business as usual. Or will be once it is refined a bit more.
Big Picture Summary
Key Findings Explained
Radiologist’s conclusion:
Both possibilities are non-cancerous
Radiologist’s Final Impression (Simplified)
One Important Note
What This Means For You
Questions You Might Have Asked (or still could)
Bottom Line
