The roundtable, entitled, “AI and Indigenous Economic Reconciliation: Powering Communities in the Digital Economy,” took place at the annual Forward Summit at on Tsuut'ina Nation.
A little consideration of how large language models work should make anyone skeptical about AI's usefulness.
I recently read the book "Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI" by Karen FAO. It contains discussion of a paper by a group of AI researchers called, “Stochastic Parrots: On the Dangers of Stochastic Parrots: Can Language Models Be Too Big
The authors argue that large language models generate convincing text by statistically predicting likely word sequences from huge datasets — not by genuinely understanding meaning, truth, or the world.
In their view, these systems can imitate language impressively while lacking comprehension.
The ChatGPT summary ends with, "The phrase itself has become shorthand for skepticism about claims that AI systems possess genuine comprehension or intelligence."
How can a system that imitates language while lacking comprehension be trusted with things like controlling air traffic or teaching students? Seems pretty risky. Methinks it's not ready for prime time.
One final point. When these stochastic parrots don't have answers they tend to come up with non factual statements. AI researchers call this hallucination. Ordinary people may call it lying.
The undue attention being paid to what is good for business, large, small, indigenous, capitalist, cooperative, seems to me to miss a couple of significantly more important points.
1. AI is being created using the 'information' scraped from the internet........which would suggest it is a form of cultural appropriation imagining that culture= intelligence. Given the range of cultural practices, not to mention the reduction of much cultural transactions under capitalist production models........that seems to be a fallacious assumption.
2. Even if we can eliminate cultural (racist, religious) bias and situate these data centres in deep space, having algorithms that know everything about what everyone thinks they know now does not equate for me with intelligence, in any real sense. Information is not intelligence. Moving information around, however rapidly, does not guarantee sound conclusions.
There are ideas, solutions, creative breakthroughs in human thinking and production that may not be a product or synthesis of anything we know now. Data Centres may well represent a 'bigger box' than any thought system we've come up with so far.........but it is still a box.
A mechanized one at that. For my money, and hopes for my grandkids.......the future deserves better than massive machines demanding massive inputs of energy.....and water, that think they can think better than organic brain/body systems.
Imagining we can mechanize intelligence, let alone consciousness seems a special kind of insanity...right up there with colonizing Mars. What's even more scary, is it may outlaw any thinking outside its own big box....for sure, it will develop the surveillance to be able to do that.
A little consideration of how large language models work should make anyone skeptical about AI's usefulness.
I recently read the book "Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI" by Karen FAO. It contains discussion of a paper by a group of AI researchers called, “Stochastic Parrots: On the Dangers of Stochastic Parrots: Can Language Models Be Too Big
In spite of my AI skepticism I asked ChatGPT to "tell me about the 'stochastic parrots' paper" I think it did a decent job and the answer can be seen at https://chatgpt.com/share/6a068809-6608-83e8-a1f2-5f78e59d98d5
I'm not an AI researcher but as far as I can see, the ChatGPT produced a decent factual account.
I find the following part damning:
-----------------------------------------------------------------------------------------------------------------------------------------------------
The authors argue that large language models generate convincing text by statistically predicting likely word sequences from huge datasets — not by genuinely understanding meaning, truth, or the world.
In their view, these systems can imitate language impressively while lacking comprehension.
----------------------------------------------------------------------------------------------------------------------------------------------------
The ChatGPT summary ends with, "The phrase itself has become shorthand for skepticism about claims that AI systems possess genuine comprehension or intelligence."
How can a system that imitates language while lacking comprehension be trusted with things like controlling air traffic or teaching students? Seems pretty risky. Methinks it's not ready for prime time.
One final point. When these stochastic parrots don't have answers they tend to come up with non factual statements. AI researchers call this hallucination. Ordinary people may call it lying.
The undue attention being paid to what is good for business, large, small, indigenous, capitalist, cooperative, seems to me to miss a couple of significantly more important points.
1. AI is being created using the 'information' scraped from the internet........which would suggest it is a form of cultural appropriation imagining that culture= intelligence. Given the range of cultural practices, not to mention the reduction of much cultural transactions under capitalist production models........that seems to be a fallacious assumption.
2. Even if we can eliminate cultural (racist, religious) bias and situate these data centres in deep space, having algorithms that know everything about what everyone thinks they know now does not equate for me with intelligence, in any real sense. Information is not intelligence. Moving information around, however rapidly, does not guarantee sound conclusions.
There are ideas, solutions, creative breakthroughs in human thinking and production that may not be a product or synthesis of anything we know now. Data Centres may well represent a 'bigger box' than any thought system we've come up with so far.........but it is still a box.
A mechanized one at that. For my money, and hopes for my grandkids.......the future deserves better than massive machines demanding massive inputs of energy.....and water, that think they can think better than organic brain/body systems.
Imagining we can mechanize intelligence, let alone consciousness seems a special kind of insanity...right up there with colonizing Mars. What's even more scary, is it may outlaw any thinking outside its own big box....for sure, it will develop the surveillance to be able to do that.