After a few years on FB and insta -I took a break then found I’ve been unable to get back on. I get an icky feeling and lost all desire to participate. This is basically the only level of social media I participate in now. Posting on articles. I wondered what would be the next big tech thing -the thing I will fall behind on. AI is it. I do not fully grasp all the future implications and uses. I guess I lack imagination. I don’t care really. I prefer the real world more and more each passing year. Face to face communication, time with my family, being outside, petting my dog. I worry about my 20 yr old and her future. I’ve been treading water daily for an hour. Rehabbing from an accident. No screens, no music, just in the water with my thoughts. Will I ever be a grandma? Will my daughter get a good job? Will she get married? Or will she tell us one day she’s in love with an AI being? Everything gets weirder by the day.
This is why I love backpacking. I am in Canada and so you lose cell coverage. Basically it is a complete break from technology, no one can reach you, no steady drumbeat of negative news.
Yeah, it's great to out of range of a cell phone for a week. Can do this in California as well. Santa Rosa Island off the coast of Santa Barbara is mostly out of range of cell phones. Nothing but the ocean, the wind, the Torrey Pines and the Island Foxes.
Don't laugh about your daughter. I've been following the case of Replika, a chatbot virtual girlfriend/boyfriend. Apparently, the chatbot was capable of VERY sexually explicit conversations so the creator removed that ability and people went crazy accusing the company of "lobotomizing my lover" and other equally ridiculous things.
I will say chatbots can be fun (except when you're trying to get customer service) as long as they're not taken so seriously.
Think of Replika as a form of mental fly paper. All these messed up people getting stuck to it instead of getting into meaningful relationships with real human beings and reproducing. I guess the question is, does it tend to attract a particular kind of fly?
I spent years in the IT field and built some of the first generation of Internet connected apps. I've kept my hand in the field studying machine learning and robotics in my free time. This guy makes sense. There are way too many people who are either wildly optimistic or pessimistic about machine learning. That is the term people in the field prefer -- rather than AI -- since anyone who has experience behind the scenes on these things pretty much knows they're not "intelligent" in any meaningful sense.
Good journalism. Solid. That's what makes this place worth subscribing too and the New York Times not.
No, they're not "intelligent." And, yes, at my company, we've de facto "canceled" terms like "artificial intelligence" and "machine intelligence" -- there's too much connotation of Terminator and HAL 9000. We instead use "machine learning" and "automated inference" (my favorite).
Traditional computer programming is "automated deduction" -- a deductive logic machine with rules and data as inputs and transformed data as output. Machine learning starts with a general framework for associative inference and consumes vast mounds of data to refine and enhance the associative inferences to the point where they become pretty good (highly probable) rules. It's "automated induction" instead of "automated deduction" -- Aristotle, call your office.
I've never seen it described that way, but I really like the deduction / induction distinction. Based on my own small level of experience with these systems, it feels very accurate.
DA's last comment nails it - and should be posted on every wall:
"Be critical, be doubtful, get outside your narrative bubbles, and resist viral attention traps."
Find a news source that exposes you to things you'd rather not read - that disagrees, politely with your basic ideas - and forces you to reexamine them.
Better yet, if you are lucky enough to meet a person who disagrees with you, but can have a civil conversation and explore the differences. Take the time to build a relationship with that person - it will be worth the effort.
Perhaps the real positive of AI will be a reawakening of critical thought? We are all capable of it - but it requires exercise.
Dear Bari, Thanks to you and your staff. Once again you hit the nail on the head.
It could be, if it's allowed to control things, especially a large network of things. Right now, its control abilities are limited to specific circumstances, like driver assist in your car. Let's keep it that way.
Anyone who's played with a mission statement generator sees through the language construction workings. Beyond that, what is ChatGPT but anything more than a fancy 'grep' command with really good output reformatting.
Current ChatGPT can provide amazingly good output, and surprisingly absurd results at the same time. Even a poor rhetorician such as I, can lead ChatGPT into amusing contradictions.
Whilst not a keen student of "Extraordinary Popular Delusions and the Madness of Crowds", I do recall even the great thinker Thomas Edison predicted great danger from supplying the newfangled Alternating Current to households. Fast forward 100 years, and we can't imagine living without it. But then again, what do I know, I've only survived the past dozen or so existential crises.
I don't think Edison really believed that - he was just bad-mouthing the competition. And talk about getting something wrong. Just think had he succeeded. It's VHS vs Betamax all over again.
Yeah, probably it won't start a nuclear war - but I am going to go out on a limb here and say that fully autonomous vehicles probably won't exceed human driving abilities for another twenty years, if ever. Saw a convincing presentation from Danso, the company that sources electronics to Toyota and Nissan a few years ago at ISSCC. Danso is all in on driver assistance, but highly skeptical of full autonomy.
There are positive and achievable advances that can be made from AI such as computer augmented interpretation of x-rays, language translation (both written and spoken), augmented chemistry for the exploration of new drugs, and driver assistance (not fully automated driving).
From the beginning, I was highly skeptical about fully automated driving and said so at a conference in about 2017 (GOMAC). Time has only proven my skepticism to be well founded.
Still remember when someone on the board of Nvidia, in about 2002, told me that my job (analog circuit design engineering) would be replaced in a few years by "AI". Artificial intelligence has been incorporated into analog design tools, but even after twenty years, we are nowhere near replacing analog circuit designers with "AI".
To some degree, the World Economic Forum has used "AI" as cover to lay people off with the excuse that "AI" has replaced a lot of jobs. This mostly hasn't happened. Instead, a lot of jobs have been offshored to lower wage countries under the guise of "AI".
I agree with Auterbach that ChatGPT can easily be backed into a corner. I've already seen people who don't know much about AI get ChatGPT to say obviously untrue things. Seems to be quite easy to do this.
Like a lot of other big tech products, AI needs regulating. The effect of "likes" on social media sites (even substack) deserves closer attention. One thing I've been aware of for quite some time is that it is very difficult online to take the middle ground on any issue. Invariably, one ends up between camps, with people afraid to be caught in no man's land between two warring factions. A lot of important information online is suppressed because it doesn't fit into a given narrative that people are familiar with.
It's time to impose heavy duty regulation on deep fakes and some of the other obviously criminal content online.
I agree with Auerbach that something needs to be done about the tendency of social media to create information silos.
That's an excellent point -- one can already see the fallacies that ChatGPT can "replace" ... whatever ... legal research, medical research, etc. Under attentive and informed human supervision, it's a powerful and useful tool. On its own, there's zero quality control and no "common sense" checks.
I've always been a listen-to-my-gut kind of a gal. Recently I had the opportunity to talk to a super intelligent young woman who works as a coder for a new A.I. start-up company. She looked at me as if she didn't understand what I meant when I mentioned that humans, unlike A.I.'s, are able to listen to their intuitions and be guided by our internal "gut feelings." The story of Stanislav Petrov demonstrates how important it is for humans to continue listening to their "guts." I look forward to sending a copy of this article to my young coder friend! : )
Teenage is the appropriate age to read Vonnegut. I loved him at sixteen. Tried reading him at thirty and couldn't stand him anymore. Pretentious manipulative elitist.
I think Ayn Rand may be his dual. She also doesn't wear well after teenage, but is great during.
Also, I love the picture used here…I just fell off my big horse after jumping and boy do I know how it feels to not always be able to control your ride
AI is still wildly inaccurate. For example, a lawyer requested a list of legal scholars who had been accused of sexual harassment. ChatGPT reported that "Turley [a professor] had made sexually suggestive comments and attempted to touch a student while on a class trip to Alaska, citing a March 2018 article in The Washington Post as the source of the information. The problem: No such article existed. There had never been a class trip to Alaska. And Turley said he’d never been accused of harassing a student." How long can tech companies cover their butts by simply putting a disclaimer on defamatory output?
It all depends on what people do with the output. If it's an amusing parlor game, so what? If it's to become a serious research tool, obviously something needs to be done. If it's to be commercially used, it will have to be licensed for custom deployments. Just like FB and Google hoovering up and repackaging news content, royalties will have to be paid, where and when appropriate.
This is why some European countries are planning to ban ChatGPT.
Agree, it depends on how people use ChatGPT and other machine learning augmented tools, and how good the user is able to detect when the inferences are wrong.
Its always ten years away, every new humanity saving technology is always ten years away. Workable fusion power is ten years away, and has been since the 1970s, we've always been at war with East Eurasia.
To be fair, I remember as a child in the early 80's hearing a (I think) Ray Bradbury story that featured a scene where the dad read the news on a computer. That seemed absurdly futuristic at the time. Fifteen/twenty years later it was reality. In the mid-90s streaming internet video seemed a long way off—it was only 5-8 years away. I'm not dismissive of radical tech predictions any more. On the other hand, even though we have robust quadcopters, flying cars still aren't a thing... go figure.
Yes, I read this too, in the 70s. It even featured in a Star Trek book, where traveling to the past, Captain Kirk is looking at a newspaper vending machine, waiting for the headlines to change.
Yes, we will see. Kurzweil has been around a very long time and also has been very consistent. Many fellow scientists refuted his timelines for a long time but now see that he was correct and agree with him. Very interesting, and the work in his books compelling.
I keep thinking: it doesn't really matter whether the are conscious or sentient or not. If they continue to develop in ways which destroy us - unconsciously, if you will - what is the difference?
Write me a joke about Ovid’s Metamorphoses in the style of Richard Pryor.
Sure, here's a joke about Ovid's Metamorphoses in the style of Richard Pryor:
"Have you heard about Ovid's Metamorphoses? Man, that book is wild! I mean, these gods and goddesses are turning people into animals left and right. It's like an ancient version of Freaky Friday, but with way more fur and feathers. And you know what's crazy? Some of these transformations are just for petty revenge. Like, imagine getting turned into a tree because some god got mad at you for being too good-looking. That's just messed up, man. I guess the moral of the story is, be careful who you piss off, or you might end up barking up the wrong tree. Literally."
Yeah, according to the folks at OpenAI, ChatGPT is supposed to deliver non-toxic messages, but veiled threats such as "be careful who you piss off, or you might end up barking up the wrong tree" are OK.
The extended examination of "What are we doing.here?" has always submitted to and obeyed the greed and lust for power of the few who are truly addicted. I fear these tools in the hands of the security state, which is our real government. They and their corporate bedfellows are not going to pause for even a minute, much less 6 months, merely because thoughtful scientists are worried about where all this might go.
This is probably the most profoundly important thing I've read in awhile. I think that's why I liked the article yesterday about the de la Motte family. If there is one thing they have done, it's that.
After a few years on FB and insta -I took a break then found I’ve been unable to get back on. I get an icky feeling and lost all desire to participate. This is basically the only level of social media I participate in now. Posting on articles. I wondered what would be the next big tech thing -the thing I will fall behind on. AI is it. I do not fully grasp all the future implications and uses. I guess I lack imagination. I don’t care really. I prefer the real world more and more each passing year. Face to face communication, time with my family, being outside, petting my dog. I worry about my 20 yr old and her future. I’ve been treading water daily for an hour. Rehabbing from an accident. No screens, no music, just in the water with my thoughts. Will I ever be a grandma? Will my daughter get a good job? Will she get married? Or will she tell us one day she’s in love with an AI being? Everything gets weirder by the day.
This is why I love backpacking. I am in Canada and so you lose cell coverage. Basically it is a complete break from technology, no one can reach you, no steady drumbeat of negative news.
Yeah, it's great to out of range of a cell phone for a week. Can do this in California as well. Santa Rosa Island off the coast of Santa Barbara is mostly out of range of cell phones. Nothing but the ocean, the wind, the Torrey Pines and the Island Foxes.
Don't laugh about your daughter. I've been following the case of Replika, a chatbot virtual girlfriend/boyfriend. Apparently, the chatbot was capable of VERY sexually explicit conversations so the creator removed that ability and people went crazy accusing the company of "lobotomizing my lover" and other equally ridiculous things.
I will say chatbots can be fun (except when you're trying to get customer service) as long as they're not taken so seriously.
Think of Replika as a form of mental fly paper. All these messed up people getting stuck to it instead of getting into meaningful relationships with real human beings and reproducing. I guess the question is, does it tend to attract a particular kind of fly?
I spent years in the IT field and built some of the first generation of Internet connected apps. I've kept my hand in the field studying machine learning and robotics in my free time. This guy makes sense. There are way too many people who are either wildly optimistic or pessimistic about machine learning. That is the term people in the field prefer -- rather than AI -- since anyone who has experience behind the scenes on these things pretty much knows they're not "intelligent" in any meaningful sense.
Good journalism. Solid. That's what makes this place worth subscribing too and the New York Times not.
No, they're not "intelligent." And, yes, at my company, we've de facto "canceled" terms like "artificial intelligence" and "machine intelligence" -- there's too much connotation of Terminator and HAL 9000. We instead use "machine learning" and "automated inference" (my favorite).
Traditional computer programming is "automated deduction" -- a deductive logic machine with rules and data as inputs and transformed data as output. Machine learning starts with a general framework for associative inference and consumes vast mounds of data to refine and enhance the associative inferences to the point where they become pretty good (highly probable) rules. It's "automated induction" instead of "automated deduction" -- Aristotle, call your office.
I've never seen it described that way, but I really like the deduction / induction distinction. Based on my own small level of experience with these systems, it feels very accurate.
Agree on 'machine learning' being the more accurate description. Calling it AI is just a marketing gimmick. At least for now.
DA's last comment nails it - and should be posted on every wall:
"Be critical, be doubtful, get outside your narrative bubbles, and resist viral attention traps."
Find a news source that exposes you to things you'd rather not read - that disagrees, politely with your basic ideas - and forces you to reexamine them.
Better yet, if you are lucky enough to meet a person who disagrees with you, but can have a civil conversation and explore the differences. Take the time to build a relationship with that person - it will be worth the effort.
Perhaps the real positive of AI will be a reawakening of critical thought? We are all capable of it - but it requires exercise.
Dear Bari, Thanks to you and your staff. Once again you hit the nail on the head.
She first identified the nail to be hit -- that's the real value of Free Press :)
Basically follow the direction of Intel co-founder Dr. Andy Grove. Who wrote a book titled "Only The Paranoid Survive."
Be Paranoid, its apparently the key to survival in uncertain times. These days, "conspiracy theory" seems to be a code word for Forewarned.
AI is the tech world's version of gain of function research.
It could be, if it's allowed to control things, especially a large network of things. Right now, its control abilities are limited to specific circumstances, like driver assist in your car. Let's keep it that way.
Worked at DARPA MTO as a consultant between 2016 and 2018. Another amazing application of machine learning is this one:
https://www.technologyreview.com/2023/02/15/1067904/ai-automation-drug-development/
Anyone who's played with a mission statement generator sees through the language construction workings. Beyond that, what is ChatGPT but anything more than a fancy 'grep' command with really good output reformatting.
Current ChatGPT can provide amazingly good output, and surprisingly absurd results at the same time. Even a poor rhetorician such as I, can lead ChatGPT into amusing contradictions.
Whilst not a keen student of "Extraordinary Popular Delusions and the Madness of Crowds", I do recall even the great thinker Thomas Edison predicted great danger from supplying the newfangled Alternating Current to households. Fast forward 100 years, and we can't imagine living without it. But then again, what do I know, I've only survived the past dozen or so existential crises.
I don't think Edison really believed that - he was just bad-mouthing the competition. And talk about getting something wrong. Just think had he succeeded. It's VHS vs Betamax all over again.
I was an electronics tech for 40 years, yeah I know how bad it could have been.
Edison had all his patents in the DC side.
Everyone thinks AGI will start a nuclear war or vaporize people with it's brain-waves or something ... complete rubbish.
Electronic Engineer here - my analogy to video media didn't go nearly far enough as you well understand. I couldn't think of a better one though.
Yeah, probably it won't start a nuclear war - but I am going to go out on a limb here and say that fully autonomous vehicles probably won't exceed human driving abilities for another twenty years, if ever. Saw a convincing presentation from Danso, the company that sources electronics to Toyota and Nissan a few years ago at ISSCC. Danso is all in on driver assistance, but highly skeptical of full autonomy.
Very realistic comments about AI from Auerbach.
There are positive and achievable advances that can be made from AI such as computer augmented interpretation of x-rays, language translation (both written and spoken), augmented chemistry for the exploration of new drugs, and driver assistance (not fully automated driving).
From the beginning, I was highly skeptical about fully automated driving and said so at a conference in about 2017 (GOMAC). Time has only proven my skepticism to be well founded.
Still remember when someone on the board of Nvidia, in about 2002, told me that my job (analog circuit design engineering) would be replaced in a few years by "AI". Artificial intelligence has been incorporated into analog design tools, but even after twenty years, we are nowhere near replacing analog circuit designers with "AI".
To some degree, the World Economic Forum has used "AI" as cover to lay people off with the excuse that "AI" has replaced a lot of jobs. This mostly hasn't happened. Instead, a lot of jobs have been offshored to lower wage countries under the guise of "AI".
I agree with Auterbach that ChatGPT can easily be backed into a corner. I've already seen people who don't know much about AI get ChatGPT to say obviously untrue things. Seems to be quite easy to do this.
Like a lot of other big tech products, AI needs regulating. The effect of "likes" on social media sites (even substack) deserves closer attention. One thing I've been aware of for quite some time is that it is very difficult online to take the middle ground on any issue. Invariably, one ends up between camps, with people afraid to be caught in no man's land between two warring factions. A lot of important information online is suppressed because it doesn't fit into a given narrative that people are familiar with.
It's time to impose heavy duty regulation on deep fakes and some of the other obviously criminal content online.
I agree with Auerbach that something needs to be done about the tendency of social media to create information silos.
That's an excellent point -- one can already see the fallacies that ChatGPT can "replace" ... whatever ... legal research, medical research, etc. Under attentive and informed human supervision, it's a powerful and useful tool. On its own, there's zero quality control and no "common sense" checks.
I've always been a listen-to-my-gut kind of a gal. Recently I had the opportunity to talk to a super intelligent young woman who works as a coder for a new A.I. start-up company. She looked at me as if she didn't understand what I meant when I mentioned that humans, unlike A.I.'s, are able to listen to their intuitions and be guided by our internal "gut feelings." The story of Stanislav Petrov demonstrates how important it is for humans to continue listening to their "guts." I look forward to sending a copy of this article to my young coder friend! : )
Teenage is the appropriate age to read Vonnegut. I loved him at sixteen. Tried reading him at thirty and couldn't stand him anymore. Pretentious manipulative elitist.
I think Ayn Rand may be his dual. She also doesn't wear well after teenage, but is great during.
I've seen people as old as 30 "turned" by Rand, but it does usually wear off after about age 20 or 25. You get a life.
Also, I love the picture used here…I just fell off my big horse after jumping and boy do I know how it feels to not always be able to control your ride
AI is still wildly inaccurate. For example, a lawyer requested a list of legal scholars who had been accused of sexual harassment. ChatGPT reported that "Turley [a professor] had made sexually suggestive comments and attempted to touch a student while on a class trip to Alaska, citing a March 2018 article in The Washington Post as the source of the information. The problem: No such article existed. There had never been a class trip to Alaska. And Turley said he’d never been accused of harassing a student." How long can tech companies cover their butts by simply putting a disclaimer on defamatory output?
To not include citations was a questionable design choice in ChatGPT. But I see now that citations can be invented. Trust but verify.
I've heard from several people that they have found that ChapGPT cites papers that don't exist.
It all depends on what people do with the output. If it's an amusing parlor game, so what? If it's to become a serious research tool, obviously something needs to be done. If it's to be commercially used, it will have to be licensed for custom deployments. Just like FB and Google hoovering up and repackaging news content, royalties will have to be paid, where and when appropriate.
This is why some European countries are planning to ban ChatGPT.
Agree, it depends on how people use ChatGPT and other machine learning augmented tools, and how good the user is able to detect when the inferences are wrong.
"A lot of what you get out of these machines is what you put in, and these machines are not anywhere close to conscious or sentient."
lol. what do you get out of these machines that you don't put in?
The Singularity is Near and How to Create a Mind, both by Ray Kurzweil. AI will pass the Turing test by 2030.
Usually it’s 10 years away, but 7 is close. I guess we’ll see.
Its always ten years away, every new humanity saving technology is always ten years away. Workable fusion power is ten years away, and has been since the 1970s, we've always been at war with East Eurasia.
To be fair, I remember as a child in the early 80's hearing a (I think) Ray Bradbury story that featured a scene where the dad read the news on a computer. That seemed absurdly futuristic at the time. Fifteen/twenty years later it was reality. In the mid-90s streaming internet video seemed a long way off—it was only 5-8 years away. I'm not dismissive of radical tech predictions any more. On the other hand, even though we have robust quadcopters, flying cars still aren't a thing... go figure.
Yes, I read this too, in the 70s. It even featured in a Star Trek book, where traveling to the past, Captain Kirk is looking at a newspaper vending machine, waiting for the headlines to change.
Wait, we're still at war with East Eurasia?!
I suspect it's like the Club of Rome and Population Bomb. It's a sliding limit.
Yes, we will see. Kurzweil has been around a very long time and also has been very consistent. Many fellow scientists refuted his timelines for a long time but now see that he was correct and agree with him. Very interesting, and the work in his books compelling.
I keep thinking: it doesn't really matter whether the are conscious or sentient or not. If they continue to develop in ways which destroy us - unconsciously, if you will - what is the difference?
Here's what I got:
---
Hello ChatGPT
Hello! How can I assist you today?
Write me a joke about Ovid’s Metamorphoses in the style of Richard Pryor.
Sure, here's a joke about Ovid's Metamorphoses in the style of Richard Pryor:
"Have you heard about Ovid's Metamorphoses? Man, that book is wild! I mean, these gods and goddesses are turning people into animals left and right. It's like an ancient version of Freaky Friday, but with way more fur and feathers. And you know what's crazy? Some of these transformations are just for petty revenge. Like, imagine getting turned into a tree because some god got mad at you for being too good-looking. That's just messed up, man. I guess the moral of the story is, be careful who you piss off, or you might end up barking up the wrong tree. Literally."
Yeah, according to the folks at OpenAI, ChatGPT is supposed to deliver non-toxic messages, but veiled threats such as "be careful who you piss off, or you might end up barking up the wrong tree" are OK.
Well, it is a joke, not to be taken too seriously.
Yes.
: )
But I take this seriously:
https://www.digitaltrends.com/computing/chatgpt-created-malware/
The extended examination of "What are we doing.here?" has always submitted to and obeyed the greed and lust for power of the few who are truly addicted. I fear these tools in the hands of the security state, which is our real government. They and their corporate bedfellows are not going to pause for even a minute, much less 6 months, merely because thoughtful scientists are worried about where all this might go.
Great article, thank you
"resist viral attention traps".
This is probably the most profoundly important thing I've read in awhile. I think that's why I liked the article yesterday about the de la Motte family. If there is one thing they have done, it's that.