Best explanation of AI I've found, although few people understand what the dependence on probabilities implies: there will always be some hallucinations, but the frequency will decline. As an ancient programmer, no longer active, I have a concern regarding AI-written code and the potential propagation of either a subtle error or a sneaky backdoor. Those kinds of bias are easy to inject if you're in control of the AI system and its default behavior, and they're devilishly difficult to detect when you use AI to conceal them. I suspect we'll eventually have an entire hierarchy of AI systems policing other AI systems, roughly the equivalent of the Hawtch-Hawtcher Bee Watcher Bee Watcher Watcher (thank you, Dr. Seuss).
Thanks, I appreciate the thumbs up. The backdoor issue is huge. I didn't even touch upon that. We need to elect a congress willing to build a set of laws around its use, and we need to do this yesterday. I don't have any specific ideas on how those laws should look, only that they are necessary.
The probability of electing a Congress competent to even discuss AI is infinitesimal. Any attempt to constrain AI will be met by a blizzard of donations and threats to "primary" a competitor in the next election. Any possible legislation will be "words without effect." Money talks; lots of money talks very loudly...
Yup. There's no going back. Big money owns the media, the technology, and the elections. I suspect we'll see a steady stream of departures from this country by scientists, professors, tech gurus, and others with marketable skills. Alternatively, the country breaks apart. See https://roguesystemsrecon.substack.com/p/breaking-up-the-usa-systemic-factors
Oh my! I'm a 70 year-old who knows very little about how digital tech works, but like everyone else, has come to rely on it. I must admit to being wary and somewhat frightened of AI, and wish there were an easy way for me to opt out of using it. While I can see how it may be used to serve humankind in beneficial ways (medicine, unlocking seemingly intractable mysteries of science, and the like) it worries me that if in the wrong hands it can be used for horrific purposes -- never mind that it will be a huge disruptor to the workplace. I worry for my grandkids who will soon need to navigate a career path in this rapidly changing landscape.
I look forward to reading the piece, lengthy as it is, to learn if my fears are justified.
Thanks. I'm 67, and the way I look at it is that it's mostly "their" problem -- they being the younger folks. I've been dead set against its use for creative stuff because I see blatant rip offs, especially in art. But I'm afraid it's with us, so we need to start thinking about how to govern its use. These are conversations we are not having.
I agree -- it is "their" problem, but like so many others (impending climate catastrophe, mounting federal debt, likely no access to Social Security when they reach our age, shrinking of the middle class, growing inaccessibility to home ownership, and on and on...) it's one given to them by their elders. And I also (and very strongly) agree with regulation -- AI is indeed here to stay, and there must be guardrails. As is true of so much technology, AI's advances constantly outpace our ability to keep up with reasonable regulation.
Also (and apologies for being somewhat of a serial commenter here) as it happens, I picked up a copy of Michael Connelly's (The Lincoln Lawyer) 'The Proving Ground', a fictional account of a mother suing an AI company for encouraging the suicide of her child. Just getting into it these past couple nights.
"Could lyricists use it to help get a better understanding for the cadence of their song?"
For me, the answer is yes. Putting it to music really exposes the flaws in cadence.
Question for you: Say AI came up with a way to reverse the effects of a stroke without surgical intervention. Would you, as a stroke survivor, be enthusiastic about using it, in that case?
In that situation, I might not fight against using it. I'd probably be very eager.
As an aside, I wonder what would happen if I used the prompt "complex guitar solo" in a song. I should try that out.
Thanks for the kind words, as usual. I can see the lyricist thing, too. I'd be curious to know how musicians feel about it, once they get pas the initial "arrrrgh." :-)
I think the medical field is where you’ll see the best use of AI. So, yeah, probably, especially if it was non-invasive. Medicine is an area where the use of prior art can’t be a bad thing. Lessons learned, etc. My biggest complaint with AI is its use in creative endeavors. Develop LLMs for science and medicine, please. Let that be your focus. That’s what I’d say to AI companies.
That's terrific! I'm also thinking along the lines of the audio system playing Bad Bunny over the Fox talking heads. And for Noem, maybe Patti Page singing "How Much is that Doggie in the Window" on repeat, and any Al Green song whenever the bedroom door on the jet is opened. Cheers!
What a ton of work you did here--thank you. It's a deep and convoluted subject, and yep, nobody knows the actual facts, if there are any. It's fun to think about--interesting to find tolerable ways to use its capabilities. What a great read, Charles. I've watched this AI space for more than two year, but from a far less lofty position than your experience gives you.
Thanks, Maryan! That reminds me, I need to go back and finish the one that you wrote. I was in the middle of finishing this one and I think I left it hanging! Your takes on this are always insightful.
Best explanation of AI I've found, although few people understand what the dependence on probabilities implies: there will always be some hallucinations, but the frequency will decline. As an ancient programmer, no longer active, I have a concern regarding AI-written code and the potential propagation of either a subtle error or a sneaky backdoor. Those kinds of bias are easy to inject if you're in control of the AI system and its default behavior, and they're devilishly difficult to detect when you use AI to conceal them. I suspect we'll eventually have an entire hierarchy of AI systems policing other AI systems, roughly the equivalent of the Hawtch-Hawtcher Bee Watcher Bee Watcher Watcher (thank you, Dr. Seuss).
Thanks, I appreciate the thumbs up. The backdoor issue is huge. I didn't even touch upon that. We need to elect a congress willing to build a set of laws around its use, and we need to do this yesterday. I don't have any specific ideas on how those laws should look, only that they are necessary.
The probability of electing a Congress competent to even discuss AI is infinitesimal. Any attempt to constrain AI will be met by a blizzard of donations and threats to "primary" a competitor in the next election. Any possible legislation will be "words without effect." Money talks; lots of money talks very loudly...
That speaks to a more general problem — money in politics, which is one that if we don't solve will probably end us anyway.
Yup. There's no going back. Big money owns the media, the technology, and the elections. I suspect we'll see a steady stream of departures from this country by scientists, professors, tech gurus, and others with marketable skills. Alternatively, the country breaks apart. See https://roguesystemsrecon.substack.com/p/breaking-up-the-usa-systemic-factors
Oh my! I'm a 70 year-old who knows very little about how digital tech works, but like everyone else, has come to rely on it. I must admit to being wary and somewhat frightened of AI, and wish there were an easy way for me to opt out of using it. While I can see how it may be used to serve humankind in beneficial ways (medicine, unlocking seemingly intractable mysteries of science, and the like) it worries me that if in the wrong hands it can be used for horrific purposes -- never mind that it will be a huge disruptor to the workplace. I worry for my grandkids who will soon need to navigate a career path in this rapidly changing landscape.
I look forward to reading the piece, lengthy as it is, to learn if my fears are justified.
Thanks. I'm 67, and the way I look at it is that it's mostly "their" problem -- they being the younger folks. I've been dead set against its use for creative stuff because I see blatant rip offs, especially in art. But I'm afraid it's with us, so we need to start thinking about how to govern its use. These are conversations we are not having.
I agree -- it is "their" problem, but like so many others (impending climate catastrophe, mounting federal debt, likely no access to Social Security when they reach our age, shrinking of the middle class, growing inaccessibility to home ownership, and on and on...) it's one given to them by their elders. And I also (and very strongly) agree with regulation -- AI is indeed here to stay, and there must be guardrails. As is true of so much technology, AI's advances constantly outpace our ability to keep up with reasonable regulation.
You're right. We've left them quite a mess.
Also (and apologies for being somewhat of a serial commenter here) as it happens, I picked up a copy of Michael Connelly's (The Lincoln Lawyer) 'The Proving Ground', a fictional account of a mother suing an AI company for encouraging the suicide of her child. Just getting into it these past couple nights.
I'd be interested in that. I've read some of his books, and I like the TV series. :-)
Also, no apology needed. Comments are always welcome. :-)
Excellent work, Charles. 👍
"Could lyricists use it to help get a better understanding for the cadence of their song?"
For me, the answer is yes. Putting it to music really exposes the flaws in cadence.
Question for you: Say AI came up with a way to reverse the effects of a stroke without surgical intervention. Would you, as a stroke survivor, be enthusiastic about using it, in that case?
In that situation, I might not fight against using it. I'd probably be very eager.
As an aside, I wonder what would happen if I used the prompt "complex guitar solo" in a song. I should try that out.
Thanks for the kind words, as usual. I can see the lyricist thing, too. I'd be curious to know how musicians feel about it, once they get pas the initial "arrrrgh." :-)
I think the medical field is where you’ll see the best use of AI. So, yeah, probably, especially if it was non-invasive. Medicine is an area where the use of prior art can’t be a bad thing. Lessons learned, etc. My biggest complaint with AI is its use in creative endeavors. Develop LLMs for science and medicine, please. Let that be your focus. That’s what I’d say to AI companies.
I agree - that would be exponentially more useful while being less piratical of creative efforts.
Oh, and please do reprogram Humpty's Qatari AF1! Might as well do Noem's Mile High Clubjet while you're at it.
If nothing else, I can program it to hug the surface like a cruise missile to scare the little darlings who ride the things.
That's terrific! I'm also thinking along the lines of the audio system playing Bad Bunny over the Fox talking heads. And for Noem, maybe Patti Page singing "How Much is that Doggie in the Window" on repeat, and any Al Green song whenever the bedroom door on the jet is opened. Cheers!
What a ton of work you did here--thank you. It's a deep and convoluted subject, and yep, nobody knows the actual facts, if there are any. It's fun to think about--interesting to find tolerable ways to use its capabilities. What a great read, Charles. I've watched this AI space for more than two year, but from a far less lofty position than your experience gives you.
Thanks, Maryan! That reminds me, I need to go back and finish the one that you wrote. I was in the middle of finishing this one and I think I left it hanging! Your takes on this are always insightful.
Makes me nervous to have your eyes there, but you are such a well-rounded, sensible, sane man. Have at it.