AI/ML seems to be following the hype cycle of every new technology since Gutenberg's press.
Jevon's Paradox tells us we'll net more tech jobs than we lose. But those jobs may consist of work we can't even imagine today. Lotus 1-2-3 clobbered accounting clerks, but enable new jobs, like financial analysts, modelers and quants.
My day-to-day is spent helping an insurance company make AI boring (automate new business submission intake, triage and route claims, detect fraud signals earlier...yawn). The stuff AI is good at cannot justify the AI companies' valuations.
This bubble will pop. Companies will go bust. But the infrastructure will remain, and much like telecom fiber, it will be repurposed for boring, valuable functions.
A great read on this topic was just released. Django Beatty has been in this space a long time and offers some penetrating insights.
WOW! You’re telling me they spent $560 billion to make a machine that’s says “Dog Barks”! What the hell kind of return is that!?! People are losing jobs, whole careers wiped out, and for what!? For a glorified magic 8-ball that can’t land a dam plane! 30,000 Amazon workers tossed out while a chatbot explains probability with more compassion than Jeff Bezos ever did.
They spent half a trillion bucks and got a glorified autocorrect that occasionally hallucinates? That’s not intelligence. That’s like calling your parrot a philosopher because it squawked the word “Nietzsche.”
AI’s randomness is the best part. Sometimes it tells the truth, sometimes it makes things up. Just like a politician. But you know, the only difference is you can’t vote AI out of office. Not yet, anyway. And they say AI’s dangerous if it lands planes. But have you seen pilots lately? Maybe a little randomness is refreshing.
AI is just the latest religion, folks. The high priests are in Silicon Valley, the temples are the data centers, and the worshippers are you, praying that the robot won’t take your job while you’re busy asking it to write your grandma’s eulogy.
Why don't "they" ask AI what are the chances of this uncontrolled debt not causing the revaluation of gold to the dollar. Second, then ask AI what are the chances that holding crypto, cash or the 7 big tech stocks pending the possible revaluation has the highest percentage of success for the "average" investor??? It seems that AI should easily tell investors how long to stay in these investments before selling and buying "what"?? For the billions of AI cost, it certainly should be able to answer those straight forward questions.
Everyone talks about AI in art or AI for information but I've found another use.
When I started using computers and I wanted to know how to use a program I hit HELP. Then programs became too complicated and programmers kept changing programs faster than the help writers could keep up.
So now when you buy a program, you have to hope that you will be able to work out how to use it. There might be a book about it that will come out in about three years time and give you details of how to use today's program, but not the one that was updated yesterday.
I have found that ChatGPT is the answer. All you need is patience...
Say you are an expert in using Windows 10. You have the desktop set out exactly as you want Then you change to Windows 11. Suddenly the Operating System starts heaping desktops on top of each other, and you don't know how to get back to the first one.
No problem... Ask AI. Without hesitation it will tell you exactly what to do. The only trouble is that the programmers changed it yesterday, and the button that you haven't pressed is not there any more. No problem... you can send a screenshot to AI which will say something like "Perfect. I see just where you are. In version XYZ they have changed it. You no longer do it that way... in fact they no longer allow you to do it at all, but here is a workaround...
After a few hours of that, you will be ably to call up whichever version of desktop you want. Then AI will suggest "Would you like me to show you how to prevent more than one version of Desktop appearing and how to call up that version with just one click? In less than another hour you have Desktop performing just as well in Windows 11 as it used to in Windows 10.
Of course, that's an easy job - it only took half a day. I usually have to spend all day. After the Windows update on July 23rd it took AI six weeks to discover all the new firewalls etc preventing it working smoothly for me. But I couldn't have guessed any of it, and I couldn't have taken it to a technician, because he isn't a magician... he can't guess the latest changes any more than I can.
Do what I did a long time ago..... get an iMac :) did that in 1984 when it first came out and realized I wasn't smart enough to use DOS operating. easy peasy!
In my experience over just the last 10 years, the quality of products and services provided by human beings has fallen off a cliff. My life is transaction intensive, and whereas prior to Covid I might encounter 2 or 3 problem transactions out of 10, now the rate is approaching 8 out of 10.
What does that look like? It means only one in five transactions has no quality defect whatsoever, product or service. The other four have problems ranging from mild to egregious. At times this becomes exasperating. The issues are both individual(an incompetent or careless human) and increasingly, systemic(eg, online systems such as banking software that do not seem to work without major glitches). All of this emerged pre-AI. Covid was a major catalyst for everything getting markedly worse. Will AI reverse this trend or exacerbate it? Time will tell, but the bar is not a high one.
as someone who worked as aircrew for twelve years, I assure you things break on aircraft all the time. Some of them prevent you from flying that day. Some are simply written up to be fixed later. If AI is left to manage the process of flying the aircraft and managing the maintenance, that calculation of accidents would change rapidly and not in a good way. I for one would no longer board an aircraft.
I always enjoy reading your commentary, Bill, and this one is no exception. However, as a physicist, it always irritates me when people can’t get basic maths right. I think you’ll find that a postulated profit of 10% of $35 billion represents a 0.6% return on $560 billion invested, NOT 0.006%, as you stated.
So, AI is not perfect - intentionally set that way per Bill. But what if you want a little perfection in your life? That is, as close to perfection as humanly possible? The recording is okay (from a cell phone, I believe - but a good set of "ear buds" helps) The lighting could be better. But there is nothing better than these talented young women.
A small but important detail. Machines don't always get 4. If floating points are used it will depend on how the mantissa and exponent are handled. It is possible to put 4 in and have 4.000000001 back or 3.9999999999. I have seen numerous times where a mismatch occurs for this reason.
Ask your self; If I was going to add a little bit of freedom to an AI program to allow the program to add some of its own "opinion" then how would I do that?
Then after I allow the computer's decision; Is that going to be right or wrong?
Thanks for your take on this space, Bill.
AI/ML seems to be following the hype cycle of every new technology since Gutenberg's press.
Jevon's Paradox tells us we'll net more tech jobs than we lose. But those jobs may consist of work we can't even imagine today. Lotus 1-2-3 clobbered accounting clerks, but enable new jobs, like financial analysts, modelers and quants.
My day-to-day is spent helping an insurance company make AI boring (automate new business submission intake, triage and route claims, detect fraud signals earlier...yawn). The stuff AI is good at cannot justify the AI companies' valuations.
This bubble will pop. Companies will go bust. But the infrastructure will remain, and much like telecom fiber, it will be repurposed for boring, valuable functions.
A great read on this topic was just released. Django Beatty has been in this space a long time and offers some penetrating insights.
https://leanpub.com/notartificialnotintelligent
Best regards,
Ben
WOW! You’re telling me they spent $560 billion to make a machine that’s says “Dog Barks”! What the hell kind of return is that!?! People are losing jobs, whole careers wiped out, and for what!? For a glorified magic 8-ball that can’t land a dam plane! 30,000 Amazon workers tossed out while a chatbot explains probability with more compassion than Jeff Bezos ever did.
They spent half a trillion bucks and got a glorified autocorrect that occasionally hallucinates? That’s not intelligence. That’s like calling your parrot a philosopher because it squawked the word “Nietzsche.”
AI’s randomness is the best part. Sometimes it tells the truth, sometimes it makes things up. Just like a politician. But you know, the only difference is you can’t vote AI out of office. Not yet, anyway. And they say AI’s dangerous if it lands planes. But have you seen pilots lately? Maybe a little randomness is refreshing.
AI is just the latest religion, folks. The high priests are in Silicon Valley, the temples are the data centers, and the worshippers are you, praying that the robot won’t take your job while you’re busy asking it to write your grandma’s eulogy.
So. What’s the point
‘It’s a grossly overrated gimmick”.
Dear God, please...
https://x.com/profstonge/status/1983134147355820096?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1983134147355820096%7Ctwgr%5Edae92a9dc5143cadfd523a855036b5a9b3bf21c6%7Ctwcon%5Es1_&ref_url=https%3A%2F%2Fwww.whatfinger.com%2F
Thanks SE for sending that.
Why don't "they" ask AI what are the chances of this uncontrolled debt not causing the revaluation of gold to the dollar. Second, then ask AI what are the chances that holding crypto, cash or the 7 big tech stocks pending the possible revaluation has the highest percentage of success for the "average" investor??? It seems that AI should easily tell investors how long to stay in these investments before selling and buying "what"?? For the billions of AI cost, it certainly should be able to answer those straight forward questions.
Everyone talks about AI in art or AI for information but I've found another use.
When I started using computers and I wanted to know how to use a program I hit HELP. Then programs became too complicated and programmers kept changing programs faster than the help writers could keep up.
So now when you buy a program, you have to hope that you will be able to work out how to use it. There might be a book about it that will come out in about three years time and give you details of how to use today's program, but not the one that was updated yesterday.
I have found that ChatGPT is the answer. All you need is patience...
Say you are an expert in using Windows 10. You have the desktop set out exactly as you want Then you change to Windows 11. Suddenly the Operating System starts heaping desktops on top of each other, and you don't know how to get back to the first one.
No problem... Ask AI. Without hesitation it will tell you exactly what to do. The only trouble is that the programmers changed it yesterday, and the button that you haven't pressed is not there any more. No problem... you can send a screenshot to AI which will say something like "Perfect. I see just where you are. In version XYZ they have changed it. You no longer do it that way... in fact they no longer allow you to do it at all, but here is a workaround...
After a few hours of that, you will be ably to call up whichever version of desktop you want. Then AI will suggest "Would you like me to show you how to prevent more than one version of Desktop appearing and how to call up that version with just one click? In less than another hour you have Desktop performing just as well in Windows 11 as it used to in Windows 10.
Of course, that's an easy job - it only took half a day. I usually have to spend all day. After the Windows update on July 23rd it took AI six weeks to discover all the new firewalls etc preventing it working smoothly for me. But I couldn't have guessed any of it, and I couldn't have taken it to a technician, because he isn't a magician... he can't guess the latest changes any more than I can.
Ian McAllister
Do what I did a long time ago..... get an iMac :) did that in 1984 when it first came out and realized I wasn't smart enough to use DOS operating. easy peasy!
In my experience over just the last 10 years, the quality of products and services provided by human beings has fallen off a cliff. My life is transaction intensive, and whereas prior to Covid I might encounter 2 or 3 problem transactions out of 10, now the rate is approaching 8 out of 10.
What does that look like? It means only one in five transactions has no quality defect whatsoever, product or service. The other four have problems ranging from mild to egregious. At times this becomes exasperating. The issues are both individual(an incompetent or careless human) and increasingly, systemic(eg, online systems such as banking software that do not seem to work without major glitches). All of this emerged pre-AI. Covid was a major catalyst for everything getting markedly worse. Will AI reverse this trend or exacerbate it? Time will tell, but the bar is not a high one.
Was this prose created by AI?
as someone who worked as aircrew for twelve years, I assure you things break on aircraft all the time. Some of them prevent you from flying that day. Some are simply written up to be fixed later. If AI is left to manage the process of flying the aircraft and managing the maintenance, that calculation of accidents would change rapidly and not in a good way. I for one would no longer board an aircraft.
I always enjoy reading your commentary, Bill, and this one is no exception. However, as a physicist, it always irritates me when people can’t get basic maths right. I think you’ll find that a postulated profit of 10% of $35 billion represents a 0.6% return on $560 billion invested, NOT 0.006%, as you stated.
Cmon! Tomato, Tumato.. Just land the 727... Flaps.... FLAPS! Where's the runway? OMG I left my pita dish in the fridge. And I hate blue cheese..
The Landing is the Singularity:
That man and machine are co-joined in an evolved new species.
This humanoid 2.0 will be able to take on and adapt to the emotional, outside the box characteristics and be able to improve itself along the way.
ChatGPT is the embryonic seed in tech’s womb.
So, AI is not perfect - intentionally set that way per Bill. But what if you want a little perfection in your life? That is, as close to perfection as humanly possible? The recording is okay (from a cell phone, I believe - but a good set of "ear buds" helps) The lighting could be better. But there is nothing better than these talented young women.
https://www.youtube.com/watch?v=4Xp1HRAsO5g
A small but important detail. Machines don't always get 4. If floating points are used it will depend on how the mantissa and exponent are handled. It is possible to put 4 in and have 4.000000001 back or 3.9999999999. I have seen numerous times where a mismatch occurs for this reason.
How many foreign workers did Amazon employ in 2025?
https://youtu.be/e-Ecodxn5m4?si=SG47FB7JiToOt3D7
The best description of AI yet, congrats Bill !
Ask your self; If I was going to add a little bit of freedom to an AI program to allow the program to add some of its own "opinion" then how would I do that?
Then after I allow the computer's decision; Is that going to be right or wrong?