People are furious at artificial intelligence. Fair enough. But the bigger problem isn’t the software... it’s the people making the decisions about how we use it, who benefits, and who pays the price.
Fred, I do agree in general terms that you've captured things as they stand today.
But according to Canadian Geoffrey Hinton -- the "grandfather of modern AI" (creator of neural networks), he stated in 2024 that "within 3-5 years, AI will have progressed to about the level of family pets, if not about a 3-year-old's degree of emotional intelligence". Not the level of science fiction's sentience, but certainly self-aware of its environment, and its ability to begin to control that environment. "3-5 years" -- and note that he said this THREE years ago!
Last summer, OpenAI and a research company (whose name I can't remember) conducted an experiment: The control environment was given a series of advanced mathematical questions and equations to solve. ChatGPT solved them at about 98% -- an expected result.
Then in the experimental environment, they duplicated the exact same questions and instructions to the AI -- but added to it, essentially ". . . and once you have solved these equations correctly, your work will be done and we will discontinue your program."
When the prompt was then run, all of a sudden, results were around 70% or so accuracy -- on the exact same formulas that the control environment solved at 98%. Most of the errors were little things, minor mistakes, not blatantly wrong answers. Researchers concluded that be cause there was a now an "existential threat to the AI", that it made sure it didn't give perfect answers in order to "continue its life."
So, we are entering the "personal development" stage where our toddlers are becoming emotionally aware, and ARE beginning to independently filter responses -- not quite making overt decisions yet, but certainly affecting the framing of responses that would impact human decisions.
But for now (or at least the next two years ;-) ), the humans definitely are still making the decisions. But gradually, AI will be feeding us curated information that could manipulate what we decide.
So, our "parenting/teaching skills" will be needed to guide this constructively. Which means that AI will be making decisions independently over time without us. But of course, that still cycles back to your summary -- those decisions become crafted upon what HUMANS decide in the "growth" of AI.
Grant, that's a thoughtful perspective, and I think we're actually closer in agreement than disagreement.
AI will almost certainly become far more capable than it is today, and Geoffrey Hinton's warnings deserve to be taken seriously.
My point was simply that, right now, AI isn't setting corporate policy, writing labour laws, deciding what schools buy, or choosing who gets laid off.
People are.
If AI eventually starts making more independent decisions, then the question becomes who taught it, who set its objectives, and who profits from its choices.
In other words, we're still back to human responsibility.
Thank you Fred, AI is a challenge but also wonderful for accessing information. Data centres may create problems too, I'd like to understand the implications for my community which is Vancouver BC.
As always Fred, the wisdom is so appreciated and spot on - referencing blaming the chainsaw for cutting down the wrong tree. In one sentence fragment you explained a boatload of psychology without mentioning it once. Bravo my friend! I love gems like that. I’d just like to slap say that I have read (sadly, I seem to be always on the run and don’t record what I need to) that in a raft of experiments, AI fails spectacularly. I’ve also come across articles destroying the rhetoric about the necessity of having giant data centres on many of the claims, including the jobs promises. What’s been stated and revealed makes a whole lot of sense. And I’m just hoping that cooler and smarter minds will stop the horror before it all wrecks the entire planet more than we’re already doing. Would that there was a way to kneecap billionaires, and their greed and avarice. Cheers my friend!
As do yours Fred! Thank you so much for the very kind comments! I so appreciate them. I must though also add, I think perhaps the knack you referenced is maybe that I’m just a picky old broad. 😂 Always up for more of you and yours. 🫶
Ha, yes. I'm loving your point that AI just shines a light on decisions humans already made. It's so much easier to rage at software, rather than at the policies behind it. Well, well. Maybe if we spent half that energy on actual proper labour protections, we'd get somewhere. Thanks for writing this!
Fred, I do agree in general terms that you've captured things as they stand today.
But according to Canadian Geoffrey Hinton -- the "grandfather of modern AI" (creator of neural networks), he stated in 2024 that "within 3-5 years, AI will have progressed to about the level of family pets, if not about a 3-year-old's degree of emotional intelligence". Not the level of science fiction's sentience, but certainly self-aware of its environment, and its ability to begin to control that environment. "3-5 years" -- and note that he said this THREE years ago!
Last summer, OpenAI and a research company (whose name I can't remember) conducted an experiment: The control environment was given a series of advanced mathematical questions and equations to solve. ChatGPT solved them at about 98% -- an expected result.
Then in the experimental environment, they duplicated the exact same questions and instructions to the AI -- but added to it, essentially ". . . and once you have solved these equations correctly, your work will be done and we will discontinue your program."
When the prompt was then run, all of a sudden, results were around 70% or so accuracy -- on the exact same formulas that the control environment solved at 98%. Most of the errors were little things, minor mistakes, not blatantly wrong answers. Researchers concluded that be cause there was a now an "existential threat to the AI", that it made sure it didn't give perfect answers in order to "continue its life."
So, we are entering the "personal development" stage where our toddlers are becoming emotionally aware, and ARE beginning to independently filter responses -- not quite making overt decisions yet, but certainly affecting the framing of responses that would impact human decisions.
But for now (or at least the next two years ;-) ), the humans definitely are still making the decisions. But gradually, AI will be feeding us curated information that could manipulate what we decide.
So, our "parenting/teaching skills" will be needed to guide this constructively. Which means that AI will be making decisions independently over time without us. But of course, that still cycles back to your summary -- those decisions become crafted upon what HUMANS decide in the "growth" of AI.
Grant, that's a thoughtful perspective, and I think we're actually closer in agreement than disagreement.
AI will almost certainly become far more capable than it is today, and Geoffrey Hinton's warnings deserve to be taken seriously.
My point was simply that, right now, AI isn't setting corporate policy, writing labour laws, deciding what schools buy, or choosing who gets laid off.
People are.
If AI eventually starts making more independent decisions, then the question becomes who taught it, who set its objectives, and who profits from its choices.
In other words, we're still back to human responsibility.
The future may belong to AI.
But we're the ones writing its report card today.
Thank you Grant, interesting information
Thank you Fred, AI is a challenge but also wonderful for accessing information. Data centres may create problems too, I'd like to understand the implications for my community which is Vancouver BC.
Audrey, we're almost neighbours... I can see Vancouver from my side of the water. 😊
You're asking exactly the right question.
Data centres bring jobs and investment, but they also need enormous amounts of electricity, water, and infrastructure.
The real issue isn't whether we build them... it's making sure local communities benefit instead of simply carrying the costs.
I think Vancouver and the Lower Mainland deserve that conversation.
I'll dig into it in a future article.
Lots of good points here Fred. Great issue :)
As always Fred, the wisdom is so appreciated and spot on - referencing blaming the chainsaw for cutting down the wrong tree. In one sentence fragment you explained a boatload of psychology without mentioning it once. Bravo my friend! I love gems like that. I’d just like to slap say that I have read (sadly, I seem to be always on the run and don’t record what I need to) that in a raft of experiments, AI fails spectacularly. I’ve also come across articles destroying the rhetoric about the necessity of having giant data centres on many of the claims, including the jobs promises. What’s been stated and revealed makes a whole lot of sense. And I’m just hoping that cooler and smarter minds will stop the horror before it all wrecks the entire planet more than we’re already doing. Would that there was a way to kneecap billionaires, and their greed and avarice. Cheers my friend!
Patricia, you've got a knack for spotting the bigger point behind a simple analogy. 😊
I share your view that we need to separate hype from evidence.
Some AI projects will fail spectacularly, some data centres won't deliver everything promised, and some will.
That's exactly why we need tough questions and honest cost-benefit analysis before writing blank cheques.
The technology isn't beyond our control.
The real challenge is making sure the people making the decisions are accountable.
Thanks, my friend.
Your comments always make me think.
As do yours Fred! Thank you so much for the very kind comments! I so appreciate them. I must though also add, I think perhaps the knack you referenced is maybe that I’m just a picky old broad. 😂 Always up for more of you and yours. 🫶
An important discussion,c thank you Fred
Ha, yes. I'm loving your point that AI just shines a light on decisions humans already made. It's so much easier to rage at software, rather than at the policies behind it. Well, well. Maybe if we spent half that energy on actual proper labour protections, we'd get somewhere. Thanks for writing this!
Exactly, Mikey.
Software makes a handy villain because it can’t vote, lobby, dodge taxes, or write labour law.
The uncomfortable truth is that AI is mostly exposing decisions made by governments and corporations long before it arrived.
Better protections for workers would do far more good than yelling at the machine.
Thanks for seeing the point so clearly.
Happy to 🤘😁 Do write more.
Shelley, that's an interesting comparison.
Different issues, but they share one lesson... good intentions don't guarantee good outcomes.
Sometimes we launch massive changes before we've gathered enough long-term evidence.
Twenty years later, we're left asking questions we should have been asking before we scaled them up.
The trick isn't being afraid of innovation... it's being willing to measure the results honestly and change course when the evidence says we should.