I just added an edit from Josh to this piece - a really cool thing he made:
"So a mate of mine and I have put together something we think is the next best thing. We reckon it’d be great if people could identify what content is — or isn’t — created by AI. We call it Responsible AI Disclosure, or RAID. Check it out here: http://responsibleaidisclosure.com/"
This was a great read! Thank you! I can’t believe that after my semi regular rants about capitalism, combined with the current conversations around AI that I didn’t make the connection before....the dystopia we worry AI will create is actually already here and has been for ages 🤦🏻♀️ seems so clear now that I’ve read it. What a great distraction it serves to be, while we worry about what AI will become we’re not thinking about what’s already happening. It reminds me of that quote, “this is the way the world ends, not with a bang, but a whimper”. As I type, I realise I sound very conspiracy theorist-ish, I’m not saying any of this is a master plan or manipulation, more that the conditions make for a world that will slowly self destruct simply by continuing as they are.
Apr 18, 2023·edited Apr 18, 2023Liked by David Farrier
While I'm not at the level of Dr. Reid, this is my field and I've been frustrated with the people that think these systems are "creating". What they're doing is pattern matching based on wide and extensive knowledge, and maximizing for the best value. It's just that they can do it at an insane rate to the human mind, so it appears to be "thinking". (One could say humans do the same, but I choose to interpret that differently.) A good parallel example is how chess is played - the computing power is effectively playing out billions of different operations, applying the outcome to the success metrics defined in the game, and maximizing for the one option out of the billions of paths that were considered. It's trying every single thing, which I like to define as "brute force" intelligence. If I had the time to play ten billion games of chess and learn from mistakes, I'd be damn good at it as well. Plus its method of "learning" is either incorporating existing available material (of which accuracy/authority can be dubious) or playing out simulations based on coded parameters - the Japanese video shows exactly how wrong those simulations can go when parameters are missing. The simulations in chess work because of the narrow game constraints, but life has so many varied parameters that it is impossible to have accounted for all of them.
The larger challenges here are actually twofold, but combine together:
* when humans determine that the output of these systems is infallible and blindly trust it (happening today with those caught by artificial images, or those using the ChatGPT output for their letter, paper, or email), so that even with human review the output is no longer questioned and effectively becomes the authority
That's the scenario of WOPR declaring that there's an inbound nuclear missile attack from the Soviets. (WarGames, 1984) There's a lot more to all of that, but this comment is already way too long and pessimistic to so many factors.
Hello. As always when I write something for Webworm, I’ll be hanging around in the comments to answer any questions you might have. Hopefully. Thanks for reading, and to David for giving me a guest spot!
Amazing read, thank you Josh and David. That video ... wow. The humanity of Hayao Miyazaki, and knowing there are many many people who feel like he does (probably most the people in this community ❤️). Frustrating that we don't often hear from them because hype and capitalism takes up so much space.
Oh! I forgot to mention - there was meant to be a link to this in the piece, but it might have fallen out. A mate of mine and I have put together a framework for responsible AI disclosures that we call RAID. It's designed as a watermark or disclaimer that users can put on their stuff to show that it was (or, importantly, wasn't!) made with the help of AI. We're pretty proud of this, and it's licensed under Creative Commons, so others can build on what we've done. http://responsibleaidisclosure.com/
Great piece Josh. Maybe, actually, it should be renamed Artificial Intent, with emphasis on the artificial? Mathematics is behind the how it's done, but humans drive the why, the intention of the tasks and outcomes. It's like by association of the word "intelligence" somehow absolves us from shitty outcomes. The intent is all ours, just generated by AI
Here https://xkcd.com/1838/ has the perfect illustration of "This is all being done with maths. Letters are converted to numbers, and complicated statistical models can be engineered to try to “guess” target numbers. Convert those numbers back into letters and boom, you’ve got sentences."
Great read. AI is equal parts terrifying thrilling for me. But as a digital artist (and a recent one at that) what pisses me off the most is how people don’t see the ‘harmless’ AI bots that make them images as anything other than fun. They don’t see theft of our art/intellectual property as anything other than just for fun. But how would they like it if Johnny Frank in the cubicle next door took bits and pieces of everyones reports in the office, edited them together, then claimed them as his own and got the promotion that they didn’t deserve? Then when they complain, Johnny says “It was just a bit of fun. A loophole. You shoulda been more careful.”
They would be pissed.
And what about the human experience? How does an AI become successful when it can’t deliver the raw emotion or opinions that the human mind can?
It is all very crazy scary and, sadly, not going away.
I say to always go for the next advancement in tech. But when is it too far? One may never know.
Great piece, Josh! I go back and forth with how I feel about AI and this really helped to clear some things up that I have been concerned about recently! Sure it's fun to see what AI *can* generate from time to time but I think Miyazaki's thoughts about it are spot on!
A neuroscientist friend said that AI like ChatGPT is like a calculator for language. It is a tool. But you do have to have some idea of how to use it otherwise it spits out numbers (or sentences). Granted, coherent sentences but it is nowhere near conscious.
Thanks for injecting some "intelligence" into the AI conversation! I was a computer programmer way back when we were just emerging from having to use zeroes & ones (000100010 ) to "tell" a computer what to do (not a working code BTW as I only witnessed this)
Glad I missed that bit because my mind doesn't work that way, so cloaking it in more "human" language allowed me to master the still primitive instructions we gave to "tell" a computer when to add or subtract, multiply or divide, and how to recognise/match trigger words to cause certain actions. I started with HUGE machines in totally atmosphere controlled rooms, that could only do basic bookkeeping functions, to now walking around with a smart phone that is expontentially & almost mindblowingly more powerful.
So I guess my mind is attuned to how increases in tech can advance so far, as we discussed then how a computer/machine "learned" from previous iterations (or the people building them more accurately?) - as in didn't need to invent so much as improve on what others invented/developed already. We had it with early programming - someone wrote a basic programme, a customer asked for something on top so you used an existing working programme & tweaked or expanded it without the need to start from scratch (and it was already debugged & proven to work, so ....) I guess that is how I come to grips with "AI" being not a complete mystery that seemed to appear suddenly, but as an increase in computing capacity NOT intelligence.
It does concern me that "AI" has so much potential for misuse/abuse and/or accidental harm (e.g. via racial bias etc), and that certain people/organisations only care about their short-term bottom line & won't allow ethics to get in the way of a good shareholder payout etc. Hope I pass on before "AI" decides who qualifies for health care & who has to die ...?
I don’t have anything smart or interesting to say, but "It’s less smart than a pig, or a fly, or even a TERF." took me completely off-guard and elicited a solid chuckle. Good write up.
I just added an edit from Josh to this piece - a really cool thing he made:
"So a mate of mine and I have put together something we think is the next best thing. We reckon it’d be great if people could identify what content is — or isn’t — created by AI. We call it Responsible AI Disclosure, or RAID. Check it out here: http://responsibleaidisclosure.com/"
This was a great read! Thank you! I can’t believe that after my semi regular rants about capitalism, combined with the current conversations around AI that I didn’t make the connection before....the dystopia we worry AI will create is actually already here and has been for ages 🤦🏻♀️ seems so clear now that I’ve read it. What a great distraction it serves to be, while we worry about what AI will become we’re not thinking about what’s already happening. It reminds me of that quote, “this is the way the world ends, not with a bang, but a whimper”. As I type, I realise I sound very conspiracy theorist-ish, I’m not saying any of this is a master plan or manipulation, more that the conditions make for a world that will slowly self destruct simply by continuing as they are.
While I'm not at the level of Dr. Reid, this is my field and I've been frustrated with the people that think these systems are "creating". What they're doing is pattern matching based on wide and extensive knowledge, and maximizing for the best value. It's just that they can do it at an insane rate to the human mind, so it appears to be "thinking". (One could say humans do the same, but I choose to interpret that differently.) A good parallel example is how chess is played - the computing power is effectively playing out billions of different operations, applying the outcome to the success metrics defined in the game, and maximizing for the one option out of the billions of paths that were considered. It's trying every single thing, which I like to define as "brute force" intelligence. If I had the time to play ten billion games of chess and learn from mistakes, I'd be damn good at it as well. Plus its method of "learning" is either incorporating existing available material (of which accuracy/authority can be dubious) or playing out simulations based on coded parameters - the Japanese video shows exactly how wrong those simulations can go when parameters are missing. The simulations in chess work because of the narrow game constraints, but life has so many varied parameters that it is impossible to have accounted for all of them.
The larger challenges here are actually twofold, but combine together:
* when humans determine that the output of these systems is infallible and blindly trust it (happening today with those caught by artificial images, or those using the ChatGPT output for their letter, paper, or email), so that even with human review the output is no longer questioned and effectively becomes the authority
* when the SALAMI systems make up completely false data that looks authoritative, called "hallucinations" (60 Minutes just covered this with their request to write a paper on Economics - https://www.cbsnews.com/video/google-ai-artificial-intelligence-advancements-60-minutes-video-2023-04-16/)
That's the scenario of WOPR declaring that there's an inbound nuclear missile attack from the Soviets. (WarGames, 1984) There's a lot more to all of that, but this comment is already way too long and pessimistic to so many factors.
Hello. As always when I write something for Webworm, I’ll be hanging around in the comments to answer any questions you might have. Hopefully. Thanks for reading, and to David for giving me a guest spot!
Amazing read, thank you Josh and David. That video ... wow. The humanity of Hayao Miyazaki, and knowing there are many many people who feel like he does (probably most the people in this community ❤️). Frustrating that we don't often hear from them because hype and capitalism takes up so much space.
Oh! I forgot to mention - there was meant to be a link to this in the piece, but it might have fallen out. A mate of mine and I have put together a framework for responsible AI disclosures that we call RAID. It's designed as a watermark or disclaimer that users can put on their stuff to show that it was (or, importantly, wasn't!) made with the help of AI. We're pretty proud of this, and it's licensed under Creative Commons, so others can build on what we've done. http://responsibleaidisclosure.com/
Great piece Josh. Maybe, actually, it should be renamed Artificial Intent, with emphasis on the artificial? Mathematics is behind the how it's done, but humans drive the why, the intention of the tasks and outcomes. It's like by association of the word "intelligence" somehow absolves us from shitty outcomes. The intent is all ours, just generated by AI
Of course, the dead giveaway that AI generated those pictures of Trump's arrest is that they are showing him actually running.
Had he been speeding down the road on a golf cart, with law enforcement shaking their fists and trailing behind, I would have been taken in too.
Here https://xkcd.com/1838/ has the perfect illustration of "This is all being done with maths. Letters are converted to numbers, and complicated statistical models can be engineered to try to “guess” target numbers. Convert those numbers back into letters and boom, you’ve got sentences."
The video made me cry as he is so right. As always great article. AI scares the daylights out of me. Thank you.
Great read. AI is equal parts terrifying thrilling for me. But as a digital artist (and a recent one at that) what pisses me off the most is how people don’t see the ‘harmless’ AI bots that make them images as anything other than fun. They don’t see theft of our art/intellectual property as anything other than just for fun. But how would they like it if Johnny Frank in the cubicle next door took bits and pieces of everyones reports in the office, edited them together, then claimed them as his own and got the promotion that they didn’t deserve? Then when they complain, Johnny says “It was just a bit of fun. A loophole. You shoulda been more careful.”
They would be pissed.
And what about the human experience? How does an AI become successful when it can’t deliver the raw emotion or opinions that the human mind can?
It is all very crazy scary and, sadly, not going away.
I say to always go for the next advancement in tech. But when is it too far? One may never know.
E x
Great piece, Josh! I go back and forth with how I feel about AI and this really helped to clear some things up that I have been concerned about recently! Sure it's fun to see what AI *can* generate from time to time but I think Miyazaki's thoughts about it are spot on!
A neuroscientist friend said that AI like ChatGPT is like a calculator for language. It is a tool. But you do have to have some idea of how to use it otherwise it spits out numbers (or sentences). Granted, coherent sentences but it is nowhere near conscious.
Thanks for injecting some "intelligence" into the AI conversation! I was a computer programmer way back when we were just emerging from having to use zeroes & ones (000100010 ) to "tell" a computer what to do (not a working code BTW as I only witnessed this)
Glad I missed that bit because my mind doesn't work that way, so cloaking it in more "human" language allowed me to master the still primitive instructions we gave to "tell" a computer when to add or subtract, multiply or divide, and how to recognise/match trigger words to cause certain actions. I started with HUGE machines in totally atmosphere controlled rooms, that could only do basic bookkeeping functions, to now walking around with a smart phone that is expontentially & almost mindblowingly more powerful.
So I guess my mind is attuned to how increases in tech can advance so far, as we discussed then how a computer/machine "learned" from previous iterations (or the people building them more accurately?) - as in didn't need to invent so much as improve on what others invented/developed already. We had it with early programming - someone wrote a basic programme, a customer asked for something on top so you used an existing working programme & tweaked or expanded it without the need to start from scratch (and it was already debugged & proven to work, so ....) I guess that is how I come to grips with "AI" being not a complete mystery that seemed to appear suddenly, but as an increase in computing capacity NOT intelligence.
It does concern me that "AI" has so much potential for misuse/abuse and/or accidental harm (e.g. via racial bias etc), and that certain people/organisations only care about their short-term bottom line & won't allow ethics to get in the way of a good shareholder payout etc. Hope I pass on before "AI" decides who qualifies for health care & who has to die ...?
I don’t have anything smart or interesting to say, but "It’s less smart than a pig, or a fly, or even a TERF." took me completely off-guard and elicited a solid chuckle. Good write up.
Thanks for such an interesting piece. So much to think about and kudos for referencing Charlie Stross (awesome Scottish sci-fi writer who manages to weave some of this into his works - and still be entertaining). If you were responsible for sharing this article https://www.google.co.nz/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwiGyL3ei9f9AhVQ1GEKHRMDDD0QFnoECA8QAQ&url=https%3A%2F%2Fnymag.com%2Fintelligencer%2Farticle%2Fai-artificial-intelligence-chatbots-emily-m-bender.html&usg=AOvVaw2EqN_HBzU9bhh2C7Mi6YFS by a computational linguist who has some fascinating things to say on exactly why we should not be taken in by ChatGPT you are truly wonderful. If not, please do share everywhere. My boss is trying to use it for everything and won’t hear a word said against it. nice pun, huh!