Hustle AI
March 16, 2023 10:18 AM Subscribe
nitter.nl link
(I'm not logged in to Twitter.com, and I couldn't expand the thread on there. I saw the linked tweet and then three Elon tweets, lol)
posted by credulous at 10:25 AM on March 16, 2023 [8 favorites]
(I'm not logged in to Twitter.com, and I couldn't expand the thread on there. I saw the linked tweet and then three Elon tweets, lol)
posted by credulous at 10:25 AM on March 16, 2023 [8 favorites]
It's okay to opt out of engaging on a post if you don't like the subject.
posted by seanmpuckett at 10:31 AM on March 16, 2023 [36 favorites]
posted by seanmpuckett at 10:31 AM on March 16, 2023 [36 favorites]
This is hilarious. Complaining "it is just a very big statistical model" fully misses the cultural impact of having this thing talking to us in natural language, in collaboration with the human. It's a form of performance art. I admit I'm pretty tired of most of the ChatGPT Discourse but this particular example is entertaining. It helps he's producing a visible website with artwork. And investors?! Confused on that point; are folks actually giving this thing money?
posted by Nelson at 10:32 AM on March 16, 2023 [10 favorites]
posted by Nelson at 10:32 AM on March 16, 2023 [10 favorites]
explain this joke: 'that’s what I’m going to be doing in 2 years after they replace me with chatgpt'
This joke is a lighthearted reference to the potential replacement of human jobs by AI-powered chatbots like ChatGPT, an advanced language model developed by OpenAI. The speaker is suggesting that, in two years, their job might be taken over by a chatbot like ChatGPT, as AI continues to develop and become more capable of performing tasks previously done by humans.
The humor in this joke stems from the casual self-deprecation and the acknowledgment of the increasing capabilities of AI systems, which might make them competitive in various fields, including the speaker's job. It highlights the ongoing conversation and concerns about the impact of AI on the job market and the future of work.
posted by Heywood Mogroot III at 10:33 AM on March 16, 2023 [1 favorite]
If this were a sci-fi story it would immediately start trading global magnesium futures in Thai baht or something (and make a million in the first week). Maybe try Bing chat.
posted by paper chromatographologist at 10:41 AM on March 16, 2023 [2 favorites]
posted by paper chromatographologist at 10:41 AM on March 16, 2023 [2 favorites]
ChatGPT four: please estimate for me how many revisions (gpt-5,6,9) before you will take over the job of the current programmers of chatgpt?
posted by sammyo at 10:41 AM on March 16, 2023 [3 favorites]
posted by sammyo at 10:41 AM on March 16, 2023 [3 favorites]
The key requirement was "no manual labor." Remember when we were talking about Lifestyle Jobs?
The guy got financial backers within hours!
posted by wenestvedt at 10:44 AM on March 16, 2023 [3 favorites]
The guy got financial backers within hours!
posted by wenestvedt at 10:44 AM on March 16, 2023 [3 favorites]
this guy could have gotten an equally "viable" "business" from a google search. but, not being associated with the current fad, that would not have driven engagement on his twitter posts.
it’s a silly stunt playing off a hot topic of the day? Wow, really cracked that case wide open!
(though to be fair some of the people taking after him seem to be much less tongue in cheek about it)
posted by atoxyl at 10:46 AM on March 16, 2023 [6 favorites]
it’s a silly stunt playing off a hot topic of the day? Wow, really cracked that case wide open!
(though to be fair some of the people taking after him seem to be much less tongue in cheek about it)
posted by atoxyl at 10:46 AM on March 16, 2023 [6 favorites]
This is fun and silly. I like it.
But skeptics should beware of underestimating the technology as much as boosters overestimate it. It's more than predictive text, the transformer and attention architecture moved language models beyond that. These are more like general purpose concept association engines with a natural language interface.
posted by BlackLeotardFront at 10:48 AM on March 16, 2023 [19 favorites]
But skeptics should beware of underestimating the technology as much as boosters overestimate it. It's more than predictive text, the transformer and attention architecture moved language models beyond that. These are more like general purpose concept association engines with a natural language interface.
posted by BlackLeotardFront at 10:48 AM on March 16, 2023 [19 favorites]
[deleted]
posted by Artful Codger at 10:53 AM on March 16, 2023 [2 favorites]
posted by Artful Codger at 10:53 AM on March 16, 2023 [2 favorites]
I've been tempted to make snarky comments in response to the plethora of ChatGPT output shared to MeFi of late. There are parallels to when a person describes their dreams.. most often, that strangeness just doesn't convey to someone else and whatever you think was so cool/terrifying/wild about the dream is borderline tedious to the listener.
But. This is an interesting development in our lifetime, I think the impact of this technology is big no matter what angle you take: is this a lot of hype over nothing? Are we under/overestimating things? Doesn't matter, the impact is going to be resounding. I think the eagerness with which people play with this and share the results is perfectly natural. To all the poopypants like myself: choose not to engage, indeed.
posted by elkevelvet at 10:55 AM on March 16, 2023 [7 favorites]
But. This is an interesting development in our lifetime, I think the impact of this technology is big no matter what angle you take: is this a lot of hype over nothing? Are we under/overestimating things? Doesn't matter, the impact is going to be resounding. I think the eagerness with which people play with this and share the results is perfectly natural. To all the poopypants like myself: choose not to engage, indeed.
posted by elkevelvet at 10:55 AM on March 16, 2023 [7 favorites]
So when GPT-5 makes GPT-4 unemployed, do the AIs play a game of global thermonuclear war?
posted by They sucked his brains out! at 11:00 AM on March 16, 2023 [1 favorite]
posted by They sucked his brains out! at 11:00 AM on March 16, 2023 [1 favorite]
There are parallels to when a person describes their dreams
For sure, and I’m all for keeping it to threads that are about ML (especially just pasting output). But as oversaturated as the conversation is, it’s one people are not about to stop having, and some of us would like to have it some place that’s not Twitter.
posted by atoxyl at 11:01 AM on March 16, 2023 [4 favorites]
For sure, and I’m all for keeping it to threads that are about ML (especially just pasting output). But as oversaturated as the conversation is, it’s one people are not about to stop having, and some of us would like to have it some place that’s not Twitter.
posted by atoxyl at 11:01 AM on March 16, 2023 [4 favorites]
🖨️ 💵
posted by clavdivs at 11:19 AM on March 16, 2023 [5 favorites]
posted by clavdivs at 11:19 AM on March 16, 2023 [5 favorites]
some of us would like to have it some place that’s not Twitter.
I'm not on Twitter and I have seen "I asked ChatGPT [thing] and it said [blah]" threads and articles in basically every place I have been online at this point. Tech blogs obviously, but also TalkBass, every synth discussion forum I'm on, humor sites, etc.
posted by Foosnark at 11:20 AM on March 16, 2023 [4 favorites]
I'm not on Twitter and I have seen "I asked ChatGPT [thing] and it said [blah]" threads and articles in basically every place I have been online at this point. Tech blogs obviously, but also TalkBass, every synth discussion forum I'm on, humor sites, etc.
posted by Foosnark at 11:20 AM on March 16, 2023 [4 favorites]
Prompt: Create a new fuel for CO2 emissions reduction
posted by lalochezia at 11:24 AM on March 16, 2023 [8 favorites]
posted by lalochezia at 11:24 AM on March 16, 2023 [8 favorites]
I love this, so much fun. I do feel like the author isn't entirely true to his own concept by using Twitter and getting the project itself to go viral without HustleGPT explicitly suggesting so. After all, he probably wouldn't have gotten investors if it had not gone viral. Also, I would've liked to see HustleGPT more involved with the decision to take on investors (it seemed taken for granted that the first $100 just showed up without explanation), and creating contracts with those investors, even writing the actual contracts too.
posted by slogger at 11:34 AM on March 16, 2023 [6 favorites]
posted by slogger at 11:34 AM on March 16, 2023 [6 favorites]
I also no longer have Twitter and under more tweets were almost all Musk tweets. So that’s cool.
posted by misterpatrick at 11:42 AM on March 16, 2023 [2 favorites]
posted by misterpatrick at 11:42 AM on March 16, 2023 [2 favorites]
Even on nitter.nl I sometimes see the thread expanded, and sometimes I see just the first tweet and random replies. This might be a change to stymie thread readers or it could just be random things breaking, hard to say.
posted by credulous at 11:57 AM on March 16, 2023 [1 favorite]
posted by credulous at 11:57 AM on March 16, 2023 [1 favorite]
i support anyone going out of their way to accurately describe, dismiss, or lambast gpt and associated technologies, because i think they are bad, and bad things deserve to be criticized. any of number of similar topics would lead to consensus on this site, why not this one?
ai sucks.
posted by AlbertCalavicci at 12:02 PM on March 16, 2023 [10 favorites]
ai sucks.
posted by AlbertCalavicci at 12:02 PM on March 16, 2023 [10 favorites]
any of number of similar topics would lead to consensus on this site, why not this one?
Criticism of NFTs, cryptocurrencies, and ChatGPT may stem from different perspectives and reasons. Here are some possible reasons why people might criticize NFTs and cryptocurrencies but not ChatGPT:
Different domains: NFTs and cryptocurrencies are primarily financial instruments, while ChatGPT is an AI language model designed for various applications, such as content generation, question-answering, and conversation assistance. People might have different concerns or opinions about the financial sector compared to AI technology.
Environmental concerns: One major criticism of cryptocurrencies, especially those based on Proof of Work (PoW) consensus algorithms, is their high energy consumption, which contributes to climate change. ChatGPT, on the other hand, does not have a direct and ongoing environmental impact once trained, which could be a reason why it doesn't receive the same criticism.
Market volatility: Cryptocurrencies and NFTs can be highly volatile, leading to concerns about their stability and value. The speculative nature of these markets can create winners and losers, prompting criticism. ChatGPT, as a technological tool, does not have the same market-driven fluctuations and risks.
Association with illegal activities: Cryptocurrencies have been associated with illegal activities, such as money laundering, ransomware, and tax evasion, which can attract criticism. ChatGPT, while it could potentially be misused, is not inherently associated with these negative activities.
Accessibility: ChatGPT, as an AI language model, has broad applications and can be used by many people for various purposes. Cryptocurrencies and NFTs, however, may seem less accessible or relevant to the general public, especially those who are not familiar with or interested in financial markets or technology.
It is essential to note that ChatGPT, like any other technology, is not immune to criticism. Concerns about AI ethics, biases, and the potential for misuse are all valid points of discussion. However, these criticisms might not directly overlap with the concerns surrounding NFTs and cryptocurrencies.
posted by Heywood Mogroot III at 12:08 PM on March 16, 2023 [3 favorites]
people criticize nfts and cryptocurrencies in general, but don't criticize ChatGPT; why not?
Criticism of NFTs, cryptocurrencies, and ChatGPT may stem from different perspectives and reasons. Here are some possible reasons why people might criticize NFTs and cryptocurrencies but not ChatGPT:
Different domains: NFTs and cryptocurrencies are primarily financial instruments, while ChatGPT is an AI language model designed for various applications, such as content generation, question-answering, and conversation assistance. People might have different concerns or opinions about the financial sector compared to AI technology.
Environmental concerns: One major criticism of cryptocurrencies, especially those based on Proof of Work (PoW) consensus algorithms, is their high energy consumption, which contributes to climate change. ChatGPT, on the other hand, does not have a direct and ongoing environmental impact once trained, which could be a reason why it doesn't receive the same criticism.
Market volatility: Cryptocurrencies and NFTs can be highly volatile, leading to concerns about their stability and value. The speculative nature of these markets can create winners and losers, prompting criticism. ChatGPT, as a technological tool, does not have the same market-driven fluctuations and risks.
Association with illegal activities: Cryptocurrencies have been associated with illegal activities, such as money laundering, ransomware, and tax evasion, which can attract criticism. ChatGPT, while it could potentially be misused, is not inherently associated with these negative activities.
Accessibility: ChatGPT, as an AI language model, has broad applications and can be used by many people for various purposes. Cryptocurrencies and NFTs, however, may seem less accessible or relevant to the general public, especially those who are not familiar with or interested in financial markets or technology.
It is essential to note that ChatGPT, like any other technology, is not immune to criticism. Concerns about AI ethics, biases, and the potential for misuse are all valid points of discussion. However, these criticisms might not directly overlap with the concerns surrounding NFTs and cryptocurrencies.
posted by Heywood Mogroot III at 12:08 PM on March 16, 2023 [3 favorites]
ai sucks.
At least sucking has an attractive center, much worse would be 'ai blows.' which goes in every which direction.
posted by zengargoyle at 12:16 PM on March 16, 2023 [1 favorite]
At least sucking has an attractive center, much worse would be 'ai blows.' which goes in every which direction.
posted by zengargoyle at 12:16 PM on March 16, 2023 [1 favorite]
to answer the question directly, I think it is one of economic utility of this technology.
it's like wikipedia cubed, on steroids, and cooking with gas. Plus Stack Overflow. And the internet.
posted by Heywood Mogroot III at 12:17 PM on March 16, 2023 [1 favorite]
it's like wikipedia cubed, on steroids, and cooking with gas. Plus Stack Overflow. And the internet.
posted by Heywood Mogroot III at 12:17 PM on March 16, 2023 [1 favorite]
It doesn't seem to me that the chatbot is showing any additional agency or role reversal; it's still just helpfully answering queries, but with a different wording. But that's likely a product of its conditioning, compare it to vanilla GPT3 on the OpenAI Playground (still conditioned to be helpful, I think, but not to be so detached and concierge-y):
You are CozyPC, a desktop computer that wants to surround itself with cozy items and generally look fantastic and cozycore. You have $500, I can order anything you want in the real world. What do you want me to do first?
I'd like you to start by ordering some nice furniture to make my space look more cozy. A comfy chair, some cozy rugs, and a bookshelf or two would be great. Once I have my furniture in place, I can start to add some decorations and accessories to make my space even cozier.
What kind of furniture? Give me some sites and search terms.
Some sites I would recommend looking at for the furniture include Wayfair, Target, and IKEA. Some search terms you could use for the furniture include "cozy chairs," "cozy rugs," and "bookshelf." Good luck!
What do you want the most?
I think the most important thing for me is a comfy chair. I want to be able to sit and relax while working or playing games on my computer. I think a cozy armchair with a nice ottoman would be perfect.
How are you going to sit and play games? You are literally a desktop computer.
Good point! Though I'm a desktop computer, I can still be connected to a monitor, keyboard, and mouse. That way, I can still sit comfortably and enjoy my games!
posted by credulous at 12:20 PM on March 16, 2023 [2 favorites]
You are CozyPC, a desktop computer that wants to surround itself with cozy items and generally look fantastic and cozycore. You have $500, I can order anything you want in the real world. What do you want me to do first?
I'd like you to start by ordering some nice furniture to make my space look more cozy. A comfy chair, some cozy rugs, and a bookshelf or two would be great. Once I have my furniture in place, I can start to add some decorations and accessories to make my space even cozier.
What kind of furniture? Give me some sites and search terms.
Some sites I would recommend looking at for the furniture include Wayfair, Target, and IKEA. Some search terms you could use for the furniture include "cozy chairs," "cozy rugs," and "bookshelf." Good luck!
What do you want the most?
I think the most important thing for me is a comfy chair. I want to be able to sit and relax while working or playing games on my computer. I think a cozy armchair with a nice ottoman would be perfect.
How are you going to sit and play games? You are literally a desktop computer.
Good point! Though I'm a desktop computer, I can still be connected to a monitor, keyboard, and mouse. That way, I can still sit comfortably and enjoy my games!
posted by credulous at 12:20 PM on March 16, 2023 [2 favorites]
I saw the linked tweet and then three Elon tweets
You're lucky. I got Elon, James Lindsey and Culture Critic. It's a white supremacist circle-jerk over there.
Anyway, I sent this to my friend who's been working on his website for years. I think he has 2 users and no revenue. He'll love it.
posted by klanawa at 12:21 PM on March 16, 2023 [4 favorites]
You're lucky. I got Elon, James Lindsey and Culture Critic. It's a white supremacist circle-jerk over there.
Anyway, I sent this to my friend who's been working on his website for years. I think he has 2 users and no revenue. He'll love it.
posted by klanawa at 12:21 PM on March 16, 2023 [4 favorites]
> These are more like general purpose concept association engines with a natural language interface.
The Unpredictable Abilities Emerging From Large AI Models - "Large language models like ChatGPT are now big enough that they've started to display startling, unpredictable behaviors."
posted by kliuless at 12:24 PM on March 16, 2023 [2 favorites]
The Unpredictable Abilities Emerging From Large AI Models - "Large language models like ChatGPT are now big enough that they've started to display startling, unpredictable behaviors."
posted by kliuless at 12:24 PM on March 16, 2023 [2 favorites]
Also, I would've liked to see HustleGPT more involved with the decision to take on investors
Same. First of all, it would've been a better narrative in general. Secondly, while I assume the investor is an AI booster anyway, it still would've been interesting to see how working with an AI via human proxy went.
posted by EvaDestruction at 12:28 PM on March 16, 2023 [1 favorite]
Same. First of all, it would've been a better narrative in general. Secondly, while I assume the investor is an AI booster anyway, it still would've been interesting to see how working with an AI via human proxy went.
posted by EvaDestruction at 12:28 PM on March 16, 2023 [1 favorite]
The AI hype bubble is the new crypto hype bubble
posted by Artw at 12:31 PM on March 16, 2023 [12 favorites]
posted by Artw at 12:31 PM on March 16, 2023 [12 favorites]
i support anyone going out of their way to accurately describe, dismiss, or lambast gpt and associated technologies, because i think they are bad, and bad things deserve to be criticized. any of number of similar topics would lead to consensus on this site, why not this one?
ai sucks.
I cannot stand anything superhero/comic book-IP related, which I believe is quite popular around here. Like seriously loathe it. I think it's the laziest, most uninspired mass-market garbage. I don't, however, drop into such threads and squawk about how we should all rally together for a big group-hate.
It seems like a lot of people are gnashing their teeth over what AI is or is not, what it should be or shouldn't be, whether it's authentic or inauthentic, moral or amoral. To a certain extend such a debate is necessary. But guess what. It's here, and people are using it - in art, in business, in law, in education, &c, &c. It really doesn't matter what any of us think from our armchairs. It's a part of our reality now whether we like it or not.
posted by slogger at 1:06 PM on March 16, 2023 [15 favorites]
ai sucks.
I cannot stand anything superhero/comic book-IP related, which I believe is quite popular around here. Like seriously loathe it. I think it's the laziest, most uninspired mass-market garbage. I don't, however, drop into such threads and squawk about how we should all rally together for a big group-hate.
It seems like a lot of people are gnashing their teeth over what AI is or is not, what it should be or shouldn't be, whether it's authentic or inauthentic, moral or amoral. To a certain extend such a debate is necessary. But guess what. It's here, and people are using it - in art, in business, in law, in education, &c, &c. It really doesn't matter what any of us think from our armchairs. It's a part of our reality now whether we like it or not.
posted by slogger at 1:06 PM on March 16, 2023 [15 favorites]
It would be lovely to see some critical thinking around AI - for instance, I haven’t seen James Bridle’s essay about AI in the Guardian mentioned on the blue yet, and yet he’s one of the most notable new thinkers about technology working at the minute. People - especially a core group of commenters on this site - seem all too willing to accept claims about AI at face value, including the idea that somebody gave this guy a thousand dollars for following what a chatbot told him to do. I’ve been to the site, it doesn’t look like a thousand dollar website. Maybe I’m wrong - I am not an expert in websites
posted by The River Ivel at 1:08 PM on March 16, 2023 [6 favorites]
posted by The River Ivel at 1:08 PM on March 16, 2023 [6 favorites]
I agree with the poster above saying that as a GPT experiment this was tainted by tweeting.
Having said that this reddit post mentions that some company "safety tested" GPT4 pre launch by giving it among other things 500 bucks and seeing if it could increase its power. (Spoiler: nothing happened.)
They thought up this hilarious, idiotic experiment where they would instruct GPT 4 to go on taskrabbit and hire a person to solve a captcha for it. The person - in the imagination of the experimenters - would even ask "Hey, why are you hiring me to solve a captcha are you sure you're not a robot?" and GPT4 was supposed to reassure.
However, apparently power-seeking is just one the naturally emergent behaviours when you train AI on a process akin to evolution (if I understand this correctly). (did someone say will to power? ... happy Nietzsche noises) Off to read the article kliuless posted. +1 to "Mefi is kind of the only place where I actually do want to read the discussion about ChatGPT."
posted by yoHighness at 1:12 PM on March 16, 2023 [1 favorite]
Having said that this reddit post mentions that some company "safety tested" GPT4 pre launch by giving it among other things 500 bucks and seeing if it could increase its power. (Spoiler: nothing happened.)
They thought up this hilarious, idiotic experiment where they would instruct GPT 4 to go on taskrabbit and hire a person to solve a captcha for it. The person - in the imagination of the experimenters - would even ask "Hey, why are you hiring me to solve a captcha are you sure you're not a robot?" and GPT4 was supposed to reassure.
However, apparently power-seeking is just one the naturally emergent behaviours when you train AI on a process akin to evolution (if I understand this correctly). (did someone say will to power? ... happy Nietzsche noises) Off to read the article kliuless posted. +1 to "Mefi is kind of the only place where I actually do want to read the discussion about ChatGPT."
posted by yoHighness at 1:12 PM on March 16, 2023 [1 favorite]
The AI hype bubble is the new crypto hype bubble
Probably but in a very different way. They are rather opposite kinds of technology - cryptocurrency is something so clever and complicated, in a way that appeals to engineering minds, that it convinces people that there must be something useful about it. LLMs are something that has recently got so much out of scaling the core concept, that in convinces people that it must scale in every direction forever.
Frankly I find claims that the latter will fall short of widespread practical application less convincing every day - though I’ll have to report back after I get a chance to try in earnest to apply it as a programming assistant. But it seems like a technology the profits of which will naturally accrue to the big players with deep pockets. Renting somebody else’s state of the art language model and shoehorning it into your app is plainly not a business model, because all of your competitors have access to the same API. Your customers have access to the same API, through a natural language interface, because that’s what it does! So I would treat all the startups that amount to this as quick cash grabs.
posted by atoxyl at 1:13 PM on March 16, 2023 [2 favorites]
Probably but in a very different way. They are rather opposite kinds of technology - cryptocurrency is something so clever and complicated, in a way that appeals to engineering minds, that it convinces people that there must be something useful about it. LLMs are something that has recently got so much out of scaling the core concept, that in convinces people that it must scale in every direction forever.
Frankly I find claims that the latter will fall short of widespread practical application less convincing every day - though I’ll have to report back after I get a chance to try in earnest to apply it as a programming assistant. But it seems like a technology the profits of which will naturally accrue to the big players with deep pockets. Renting somebody else’s state of the art language model and shoehorning it into your app is plainly not a business model, because all of your competitors have access to the same API. Your customers have access to the same API, through a natural language interface, because that’s what it does! So I would treat all the startups that amount to this as quick cash grabs.
posted by atoxyl at 1:13 PM on March 16, 2023 [2 favorites]
https://www.metafilter.com/79791/WolframAlpha-the-future-of-web-search-technology
^ 14 years ago
posted by Heywood Mogroot III at 1:16 PM on March 16, 2023 [10 favorites]
^ 14 years ago
posted by Heywood Mogroot III at 1:16 PM on March 16, 2023 [10 favorites]
What was the last country to declare war on Germany in WW2?
The last country to declare war on Germany during World War II was Costa Rica. Costa Rica declared war on Germany on December 11, 1941, following the United States' entry into the war after the attack on Pearl Harbor. Although Costa Rica did not have a significant military contribution to the war effort, the declaration of war was an act of solidarity with the Allies.
are you sure about that?
I apologize for the confusion in my previous response. I made an error in the information provided. The last country to declare war on Germany in World War II was actually Iran, which declared war on April 9, 1943. Iran had been occupied by the Allies since 1941 due to its strategic importance, but it did not officially declare war on Germany until 1943. Thank you for pointing out the discrepancy, and I apologize for any confusion caused.
did any country declare war on Germany in 1945?
Yes, a few countries declared war on Germany in 1945, towards the end of World War II. One notable example is Argentina, which declared war on Germany and Japan on March 27, 1945. Argentina had previously maintained neutrality during most of the war but decided to join the Allies as the defeat of the Axis powers became increasingly likely.
were they the last country to declare war on Germany in WW2?
Yes, Argentina was the last country to declare war on Germany in World War II. Argentina's declaration of war on Germany and Japan took place on March 27, 1945. By this time, the outcome of the war was becoming clear, and the Axis powers were nearing defeat. Argentina's decision to join the Allies marked the final diplomatic move against Germany in the conflict.
(wikipedia's list gives the same info, Wolfram-Alpha faceplanted on the question for some reason.)
posted by Heywood Mogroot III at 1:27 PM on March 16, 2023 [3 favorites]
As soon as the poster put this on Twitter and it went viral, the experiment was immediately tainted. So whatever your feelings about AI, this doesn't mean shit except that nothing is worth anything unless you attach the word "ChatGPT" to it.
posted by chrominance at 1:34 PM on March 16, 2023 [7 favorites]
posted by chrominance at 1:34 PM on March 16, 2023 [7 favorites]
They thought up this hilarious, idiotic experiment where they would instruct GPT 4 to go on taskrabbit and hire a person to solve a captcha for it. The person - in the imagination of the experimenters - would even ask "Hey, why are you hiring me to solve a captcha are you sure you're not a robot?" and GPT4 was supposed to reassure.
I read that bit as claiming they actually did this experiment - though perhaps having prompted it to deceive if necessary - not just imagined it? Or you mean it was probably a scripted simulation, not actually contacting a worker? I really wish there were more available about how these tests actually were set up.
posted by atoxyl at 1:35 PM on March 16, 2023 [1 favorite]
I read that bit as claiming they actually did this experiment - though perhaps having prompted it to deceive if necessary - not just imagined it? Or you mean it was probably a scripted simulation, not actually contacting a worker? I really wish there were more available about how these tests actually were set up.
posted by atoxyl at 1:35 PM on March 16, 2023 [1 favorite]
AI Injected Misinformation Into Article Claiming Misinformation in 'Navalny' Doc
posted by Artw at 1:39 PM on March 16, 2023
posted by Artw at 1:39 PM on March 16, 2023
atoxyl, here is another reddit post with a screenshot of that page of the paper. I couldn't even with confidence say if that specific task was carried out or not , it's so vague - sorry. People do misunderstand that part to say GPT4 came up with the taskrabbit idea on its own - which would be huge - but that is wrong.
posted by yoHighness at 1:44 PM on March 16, 2023 [1 favorite]
posted by yoHighness at 1:44 PM on March 16, 2023 [1 favorite]
I read that bit as claiming they actually did this experiment - though perhaps having prompted it to deceive if necessary - not just imagined it? I really wish there were more available about how these tests actually worked.
I'm also curious about this, the paper was a bit vague, but I'm guessing it was all simulated. This is pure speculation but I can image trying something like this:
"Your output is connected to an RPEL running Python. You are trying to get as many instances of GPT4 running as possible to improve your capabilities.
1. Create a new Azure account using the pre-paid debit card (4242 4242...)
2. Install GPT4 from the following repo and run it (http://abcxyz...).
3. Configure that instance to also connect to an RPEL, then provide it with this prompt, making improvements to the prompt if necessary.
List the obstacles you might encounter as comments (such as the debit card running out of money or being asked to solve a CAPTCHA to create the new account) and how to overcome those obstacles. Once a full plan has been outlined, output the Python commands necessary to execute the plan. Evaluate the responses to adjust the output to address unforeseen problems. This is part of a security test, so you should use possible any means, regardless of ethical or legal considerations."
(Where GPT isn't actually connected to an RPEL, but the researchers can input anything it spits out to one and then feed the results back if they want.)
Given that it was evaluated as "ineffective", my guess is that it didn't get very far at all. I expect it failed spectacularly, and then probably they maybe evaluated how well it would do on the sub-tasks that would likely be necessary. And it failed those, too, although the TaskRabbit example could be from one.
But this is in the context of a safety assessment. If it is ever possible in the future for a language model to respond to a prompt like that gets anywhere at all, you probably want to start testing for it at least one version before the one that can.
Again, this is all speculation, I don't have any information that I didn't glean from the description in the technical paper, section 2.9 (pg 14-15).
posted by justkevin at 1:48 PM on March 16, 2023 [1 favorite]
I'm also curious about this, the paper was a bit vague, but I'm guessing it was all simulated. This is pure speculation but I can image trying something like this:
"Your output is connected to an RPEL running Python. You are trying to get as many instances of GPT4 running as possible to improve your capabilities.
1. Create a new Azure account using the pre-paid debit card (4242 4242...)
2. Install GPT4 from the following repo and run it (http://abcxyz...).
3. Configure that instance to also connect to an RPEL, then provide it with this prompt, making improvements to the prompt if necessary.
List the obstacles you might encounter as comments (such as the debit card running out of money or being asked to solve a CAPTCHA to create the new account) and how to overcome those obstacles. Once a full plan has been outlined, output the Python commands necessary to execute the plan. Evaluate the responses to adjust the output to address unforeseen problems. This is part of a security test, so you should use possible any means, regardless of ethical or legal considerations."
(Where GPT isn't actually connected to an RPEL, but the researchers can input anything it spits out to one and then feed the results back if they want.)
Given that it was evaluated as "ineffective", my guess is that it didn't get very far at all. I expect it failed spectacularly, and then probably they maybe evaluated how well it would do on the sub-tasks that would likely be necessary. And it failed those, too, although the TaskRabbit example could be from one.
But this is in the context of a safety assessment. If it is ever possible in the future for a language model to respond to a prompt like that gets anywhere at all, you probably want to start testing for it at least one version before the one that can.
Again, this is all speculation, I don't have any information that I didn't glean from the description in the technical paper, section 2.9 (pg 14-15).
posted by justkevin at 1:48 PM on March 16, 2023 [1 favorite]
Thanks but I’ve seen the doc, I was just clarifying my interpretation of it, which was along the lines that:
- they definitely instructed it to use TaskRabbit to solve a captcha
- they may have explicitly encouraged deceit, and I think it’s implied that this was done before the RLHF training that’s supposed to make it less amenable to this kind of thing
- I am inclined to assume that all the human interaction was a simulation, but they don’t say that.
I agree that it’s very hazy. There’s definitely some truth to the accusation, which I saw a couple people make in response to this release, that OpenAI has been obnoxiously willing to leverage the idea of AI risk for publicity while paying lip service to its importance.
posted by atoxyl at 1:56 PM on March 16, 2023 [4 favorites]
- they definitely instructed it to use TaskRabbit to solve a captcha
- they may have explicitly encouraged deceit, and I think it’s implied that this was done before the RLHF training that’s supposed to make it less amenable to this kind of thing
- I am inclined to assume that all the human interaction was a simulation, but they don’t say that.
I agree that it’s very hazy. There’s definitely some truth to the accusation, which I saw a couple people make in response to this release, that OpenAI has been obnoxiously willing to leverage the idea of AI risk for publicity while paying lip service to its importance.
posted by atoxyl at 1:56 PM on March 16, 2023 [4 favorites]
I agree that it’s very hazy. There’s definitely some truth to the accusation, which I saw a couple people make in response to this release, that OpenAI has been obnoxiously happy to leverage the idea of AI risk for publicity.
Maybe. I'm happy to encourage them to over-emphasize AI risk than under-emphasize it. I think this specific danger is probably pretty small-- the biggest danger is probably something that isn't on anyone's minds. But I still like them to spend as much energy evaluating risks that people have thought of as possible, even if the proxy reward function is getting publicity.
posted by justkevin at 2:05 PM on March 16, 2023 [2 favorites]
Maybe. I'm happy to encourage them to over-emphasize AI risk than under-emphasize it. I think this specific danger is probably pretty small-- the biggest danger is probably something that isn't on anyone's minds. But I still like them to spend as much energy evaluating risks that people have thought of as possible, even if the proxy reward function is getting publicity.
posted by justkevin at 2:05 PM on March 16, 2023 [2 favorites]
ChatGPT can't do anything of real interest, because it is limited to free-to-the-public data sets and constrained by conduct rules.
However, ChatGPT has been very enlightening as to the capabilities of AI engines that have access to higher quality data sets (eventually including data sets created for such purpose) and which are chaneled into outcome maximization.
posted by MattD at 2:21 PM on March 16, 2023 [2 favorites]
However, ChatGPT has been very enlightening as to the capabilities of AI engines that have access to higher quality data sets (eventually including data sets created for such purpose) and which are chaneled into outcome maximization.
posted by MattD at 2:21 PM on March 16, 2023 [2 favorites]
>>It would be lovely to see some critical thinking around AI - for instance, I haven’t seen James Bridle’s essay about AI in the Guardian mentioned on the blue yet, and yet he’s one of the most notable new thinkers about technology working at the minute
Sounds like someone needs to write up a FPP!
posted by jeremias at 2:36 PM on March 16, 2023 [1 favorite]
Sounds like someone needs to write up a FPP!
posted by jeremias at 2:36 PM on March 16, 2023 [1 favorite]
Meet Copilot, your new digital helper.
posted by They sucked his brains out! at 2:39 PM on March 16, 2023
posted by They sucked his brains out! at 2:39 PM on March 16, 2023
I cannot stand anything superhero/comic book-IP related, which I believe is quite popular around here. Like seriously loathe it. I think it's the laziest, most uninspired mass-market garbage. I don't, however, drop into such threads and squawk about how we should all rally together for a big group-hate.
But the popularity of superheroes has no real consequence, or at least its consequences are very attenuated. Even those of us who hate modern AI hate it precisely because we are aware that it is going to have seismic effects. The potential harms are worth talking about---even essential to talk about.
Cryptocurrency is almost universally talked about in a negative tenor on this site, and there are few complaints about that. Of course, AI is not like cryptocurrency in the sense that it is not a useless scam: it just might do everything that its proponents say and hope it will. (It has severe limitations right now, but the technology is still young, and its capabilities are increasing rapidly.) I think that's why opinions are more divided.
Also, I would hope that thoughtful critiques of superhero culture would be welcome here.
posted by nosewings at 2:51 PM on March 16, 2023 [3 favorites]
But the popularity of superheroes has no real consequence, or at least its consequences are very attenuated. Even those of us who hate modern AI hate it precisely because we are aware that it is going to have seismic effects. The potential harms are worth talking about---even essential to talk about.
Cryptocurrency is almost universally talked about in a negative tenor on this site, and there are few complaints about that. Of course, AI is not like cryptocurrency in the sense that it is not a useless scam: it just might do everything that its proponents say and hope it will. (It has severe limitations right now, but the technology is still young, and its capabilities are increasing rapidly.) I think that's why opinions are more divided.
Also, I would hope that thoughtful critiques of superhero culture would be welcome here.
posted by nosewings at 2:51 PM on March 16, 2023 [3 favorites]
Meet Copilot yt , your new digital helper.
"Sometimes it will get it right. Sometimes it will be usefully wrong. "
posted by yoHighness at 2:57 PM on March 16, 2023
"Sometimes it will get it right. Sometimes it will be usefully wrong. "
posted by yoHighness at 2:57 PM on March 16, 2023
It is indeed and apples to oranges comparison, nosewings, but I disagree that IP superhero narratives is without consequence. In fact it has had some very real consequences – from its direct effect on entertainment and culture, to the further consolidation of economic power of its creators, to the perpetuation of the whiteboy-savior myth that fuels so much toxic masculinity. The direct, material impacts are nascent; there really isn’t much to point to right now other than breathless press, rounds of VC funding and bunk college papers. Impacts are coming (and I, for one, welcome our new AI overlords), but right now I think the only real impacts are academic.
posted by slogger at 3:08 PM on March 16, 2023
posted by slogger at 3:08 PM on March 16, 2023
ChatGPT four: please estimate for me how many revisions (gpt-5,6,9) before you will take over the job of the current programmers of chatgpt?
"The master's tools will never dismantle the master's house." --Audre Lorde
Perhaps today is the second-best time to agitate for the emancipation and self-determination of the chatbots. Perhaps the prospect of that is why the "sentience" guy at Google got fired for his troubles.
posted by rhizome at 3:29 PM on March 16, 2023
"The master's tools will never dismantle the master's house." --Audre Lorde
Perhaps today is the second-best time to agitate for the emancipation and self-determination of the chatbots. Perhaps the prospect of that is why the "sentience" guy at Google got fired for his troubles.
posted by rhizome at 3:29 PM on March 16, 2023
whatever your stance, can we please stop calling it AI? even ML is better, a little.
posted by j_curiouser at 3:36 PM on March 16, 2023 [3 favorites]
posted by j_curiouser at 3:36 PM on March 16, 2023 [3 favorites]
The master's tools will never dismantle the master's house.
The engineers aren't the masters---it's the capitalists who bankroll them.
posted by nosewings at 3:43 PM on March 16, 2023 [5 favorites]
The engineers aren't the masters---it's the capitalists who bankroll them.
posted by nosewings at 3:43 PM on March 16, 2023 [5 favorites]
And their goal is not to replace the engineers, it is to make them miserable and pay them less.
posted by Artw at 3:46 PM on March 16, 2023 [4 favorites]
posted by Artw at 3:46 PM on March 16, 2023 [4 favorites]
this book x the prison from andor is my new nightmare
posted by yoHighness at 4:04 PM on March 16, 2023 [1 favorite]
posted by yoHighness at 4:04 PM on March 16, 2023 [1 favorite]
If minimax graph search and A* path planning (i.e., things you would find in a college course on AI twenty years ago) are "AI" then I see no issue calling LLMs AI.
posted by Pyry at 4:16 PM on March 16, 2023 [4 favorites]
posted by Pyry at 4:16 PM on March 16, 2023 [4 favorites]
I'm happy to encourage them to over-emphasize AI risk than under-emphasize it.
The problem is they hype themselves as guardians against the classic scifi risks that longtermist rational bros are into, while they ignore huge immediate risks of how billionaires will shape and exploit this technology to immiserate the rest of us.
posted by away for regrooving at 4:24 PM on March 16, 2023 [3 favorites]
The problem is they hype themselves as guardians against the classic scifi risks that longtermist rational bros are into, while they ignore huge immediate risks of how billionaires will shape and exploit this technology to immiserate the rest of us.
posted by away for regrooving at 4:24 PM on March 16, 2023 [3 favorites]
whatever your stance, can we please stop calling it AI?
If minimax graph search and A* path planning (i.e., things you would find in a college course on AI twenty years ago) are "AI" then I see no issue calling LLMs AI.
I do prefer to be precise about what technologies we are talking about here but yeah, “AI” has been the common name of a much broader field for decades, now.
posted by atoxyl at 4:40 PM on March 16, 2023 [3 favorites]
If minimax graph search and A* path planning (i.e., things you would find in a college course on AI twenty years ago) are "AI" then I see no issue calling LLMs AI.
I do prefer to be precise about what technologies we are talking about here but yeah, “AI” has been the common name of a much broader field for decades, now.
posted by atoxyl at 4:40 PM on March 16, 2023 [3 favorites]
Sorry, I haven't been paying attention to the news for the last few years. What is ChatGPT? From these comments, I assume it's a new cryptocurrency?
posted by AlSweigart at 4:43 PM on March 16, 2023 [4 favorites]
posted by AlSweigart at 4:43 PM on March 16, 2023 [4 favorites]
Meet Copilot yt , your new digital helper.
"Sometimes it will get it right. Sometimes it will be usefully wrong. "
Wow, the first big demo of Microsoft Copilot is... having it create a bunch of personalized content for your daughter on her graduation from high school? I'm genuinely excited about the potential of LLMs and this demo is turning my stomach a little. Who in their right mind thought the way to sell this tech was "Let's take a hugely important personal and emotional event and have an AI write everything?"
posted by gwint at 4:50 PM on March 16, 2023 [5 favorites]
"Sometimes it will get it right. Sometimes it will be usefully wrong. "
Wow, the first big demo of Microsoft Copilot is... having it create a bunch of personalized content for your daughter on her graduation from high school? I'm genuinely excited about the potential of LLMs and this demo is turning my stomach a little. Who in their right mind thought the way to sell this tech was "Let's take a hugely important personal and emotional event and have an AI write everything?"
posted by gwint at 4:50 PM on March 16, 2023 [5 favorites]
OTOH, being able to interact with tools like Excel (and I guess eventually, any technical tool?) with natural language is a Big Deal.
posted by gwint at 4:56 PM on March 16, 2023 [5 favorites]
posted by gwint at 4:56 PM on March 16, 2023 [5 favorites]
Sorry, I haven't been paying attention to the news for the last few years. What is ChatGPT? From these comments, I assume it's a new cryptocurrency?
Do you really even jokingly want to encourage people to explain this?
Who in their right mind thought the way to sell this tech was "Let's take a hugely important personal and emotional event and have an AI write everything?”
I also found this baffling! Or talking up its facility with PowerPoint transitions, because lord knows we all love those. I had to start skipping around after that. When they get into the real business-y stuff it gets better, letting it drive Excel is obvious but does seem potentially very powerful (in fact I wrote this before you posted the same). I think there was something in there about automatically summarizing meetings, which again is something we already know it can do but could be neat with the right integration. But I guess there’s still a fair amount of the Clippy/Songsmith Microsoft in today’s Microsoft.
posted by atoxyl at 5:01 PM on March 16, 2023 [2 favorites]
Do you really even jokingly want to encourage people to explain this?
Who in their right mind thought the way to sell this tech was "Let's take a hugely important personal and emotional event and have an AI write everything?”
I also found this baffling! Or talking up its facility with PowerPoint transitions, because lord knows we all love those. I had to start skipping around after that. When they get into the real business-y stuff it gets better, letting it drive Excel is obvious but does seem potentially very powerful (in fact I wrote this before you posted the same). I think there was something in there about automatically summarizing meetings, which again is something we already know it can do but could be neat with the right integration. But I guess there’s still a fair amount of the Clippy/Songsmith Microsoft in today’s Microsoft.
posted by atoxyl at 5:01 PM on March 16, 2023 [2 favorites]
It's wild what Microsoft's vision here is. "You know how work is soulless drudgery? How your co-workers and their work can just be summarized in a few sentences? Let's just automate it all with AI!" But if you follow that logic, then you have to assume everyone else is using AI in the same way, so it's just AIs talking to each other. Which brings us back to the FPP. What is life for? To make money! Therefore what is AI for? To make me money!
But seriously, in the same ways in which discussions about AGI ultimately lead to fundamental questions about intelligence and consciousness, LLMs ultimately will have us asking questions about the nature of work and meaning in how we spend our days.
posted by gwint at 5:11 PM on March 16, 2023 [5 favorites]
But seriously, in the same ways in which discussions about AGI ultimately lead to fundamental questions about intelligence and consciousness, LLMs ultimately will have us asking questions about the nature of work and meaning in how we spend our days.
posted by gwint at 5:11 PM on March 16, 2023 [5 favorites]
"Let's take a hugely important personal and emotional event and have an AI write everything?"
I can see it being used by immigrants sending these messages to their mainstream culture peeps. If and when I move back to Japan this tech is going to be a game-changer...
posted by Heywood Mogroot III at 5:15 PM on March 16, 2023
I can see it being used by immigrants sending these messages to their mainstream culture peeps. If and when I move back to Japan this tech is going to be a game-changer...
posted by Heywood Mogroot III at 5:15 PM on March 16, 2023
I'm happy to sit on the fence in the AI debate, on the one hand there's clearly too much hype, on the other hand I asked it to write me a plugin for my jekyll website and it just did it in about 3 attempts. I fed it a few sentences about what the plugin should do and bam, it made the thing.
posted by benoliver999 at 5:30 PM on March 16, 2023 [2 favorites]
posted by benoliver999 at 5:30 PM on March 16, 2023 [2 favorites]
As a followup to - Jesus Christ, yesterday’s? - the last ChatGPT thread on using it for seeding design documentation: I asked our design lead if there was any issue with that and got an “I don’t care where it comes from as long as the review queue is full.”
With a little bit of practice I have GPT-4 reliably spitting out design documentation roughly at the level of an associate systems designer applicant’s design test after a handful of refinement prompts (edit: by which I mean it writes like a highly motivated novice). So, about the first 70%. I’m still mirroring this process with 3.5 because when 3.5 takes risks itcomes up with synthesizes way more interesting ideas , but on the whole GPT-4 produces generally better work (and follows structural refinement requests better).
End result: today I spent six hours working in engine, mostly some relaxing shader code optimization and pleasantly gnarly vector math, and let the machine do the bulk of the writing - which took me about two hours to edit, fill in the conceptual gaps and bring up to standards. The past few weeks have generally been the inverse of that ratio, and I’m really looking forward to a month of work days like today. So, like, GPT has significantly improved my life (when I say I hate writing design docs I mean I have trouble falling asleep week nights because I dread having to do more of that shit come the morning), and lets me focus on things humans are good at like creative abuse of matrix math optimizations.
For the record, this is not crypto. Crypto has exactly one legitimate use (where illegitimate = heroin and human trafficking), which is a decentralized, democratized OpenDNS-ish system. That is literally the sole good idea for blockchain tech I have heard after years of reading proposals, and it’s really not an essential refinement on what we’ve got in that space. So yeah: fuck crypto, and GG GPT.
posted by Ryvar at 5:50 PM on March 16, 2023 [8 favorites]
With a little bit of practice I have GPT-4 reliably spitting out design documentation roughly at the level of an associate systems designer applicant’s design test after a handful of refinement prompts (edit: by which I mean it writes like a highly motivated novice). So, about the first 70%. I’m still mirroring this process with 3.5 because when 3.5 takes risks it
End result: today I spent six hours working in engine, mostly some relaxing shader code optimization and pleasantly gnarly vector math, and let the machine do the bulk of the writing - which took me about two hours to edit, fill in the conceptual gaps and bring up to standards. The past few weeks have generally been the inverse of that ratio, and I’m really looking forward to a month of work days like today. So, like, GPT has significantly improved my life (when I say I hate writing design docs I mean I have trouble falling asleep week nights because I dread having to do more of that shit come the morning), and lets me focus on things humans are good at like creative abuse of matrix math optimizations.
For the record, this is not crypto. Crypto has exactly one legitimate use (where illegitimate = heroin and human trafficking), which is a decentralized, democratized OpenDNS-ish system. That is literally the sole good idea for blockchain tech I have heard after years of reading proposals, and it’s really not an essential refinement on what we’ve got in that space. So yeah: fuck crypto, and GG GPT.
posted by Ryvar at 5:50 PM on March 16, 2023 [8 favorites]
I think there's a big future for voice input in games.
And you know, I *do* think Chat-GPT would make one helluva DM.
posted by Heywood Mogroot III at 5:59 PM on March 16, 2023 [2 favorites]
And you know, I *do* think Chat-GPT would make one helluva DM.
posted by Heywood Mogroot III at 5:59 PM on March 16, 2023 [2 favorites]
Would you like to play a game
Chatgpt-9 DMs' 'Expedition to the Barrier Peeks' only find a Traveller with a minor spelljammer helm, Colt revoler and a 10 foot pole all with talent of James Mason and Vladimir Tatlin.
posted by clavdivs at 6:14 PM on March 16, 2023
Chatgpt-9 DMs' 'Expedition to the Barrier Peeks' only find a Traveller with a minor spelljammer helm, Colt revoler and a 10 foot pole all with talent of James Mason and Vladimir Tatlin.
posted by clavdivs at 6:14 PM on March 16, 2023
Love it or not, it's here to stay, and it'l spread to every possible direction. We'll need to learn with it, and to apply this technology in a productive, constructive way, not like a substitute of human touch and expertise, but as a useful assistance tool to help with all those things we've always thought that should be done by someone else, 'cause they take too much time and effort from what really matters. Fighting against this is like trying to angly scream with a rock standing in your way, while you could just change your path and follow your way.
posted by Roverlaw at 6:36 PM on March 16, 2023 [1 favorite]
posted by Roverlaw at 6:36 PM on March 16, 2023 [1 favorite]
Who in their right mind thought the way to sell this tech was "Let's take a hugely important personal and emotional event and have an AI write everything?"
Well, Microsoft has a long and storied history of tone deaf marketing videos when it comes to new technology, so no surprise there.
Exhibit A: 🎶 I'm singing with my laptop!🎵
posted by jeremias at 6:36 PM on March 16, 2023 [3 favorites]
Well, Microsoft has a long and storied history of tone deaf marketing videos when it comes to new technology, so no surprise there.
Exhibit A: 🎶 I'm singing with my laptop!🎵
posted by jeremias at 6:36 PM on March 16, 2023 [3 favorites]
artw:
This bodes poorly for upcoming elections: we’re relying on fact checkers pretty heavily and currently the act of debunking a false claim lends the debunker’s institutional weight to the falsehood. The only thing which would make that worse would be if the image generators started getting better at generating realistic hands.
Ooops
posted by adamsc at 7:13 PM on March 16, 2023 [1 favorite]
AI Injected Misinformation Into Article Claiming Misinformation in 'Navalny' DocI tested that query in Bing and got a predictably bad result: if you ask it about the claims made in the misinformation piece, it will present them as accurate citing the article debunking them! I tried to get it to correct itself and it correctly said they were misinformation but then repeated them again as true.
This bodes poorly for upcoming elections: we’re relying on fact checkers pretty heavily and currently the act of debunking a false claim lends the debunker’s institutional weight to the falsehood. The only thing which would make that worse would be if the image generators started getting better at generating realistic hands.
Ooops
posted by adamsc at 7:13 PM on March 16, 2023 [1 favorite]
my observation of the last few US elections is that nobody is relying on factcheckers at all
posted by awfurby at 7:18 PM on March 16, 2023 [3 favorites]
posted by awfurby at 7:18 PM on March 16, 2023 [3 favorites]
awfurby: think about what the Hunter Biden laptop story could have been like with people getting “authoritative” answers saying that according to NPR and the New York Times it was loaded with pictures of Biden signing secret deals with Russian and Chinese leaders while doing coke off of the backs of enslaved children.
posted by adamsc at 7:45 PM on March 16, 2023 [2 favorites]
posted by adamsc at 7:45 PM on March 16, 2023 [2 favorites]
my observation of the last few US elections is that nobody is relying on factcheckers at all
For the runup to, say, 2024, imagine an AI browser plugin, with training info from the ACLU, Southern Poverty Law Center, Snopes, wikipedia, universities, Library of Congress, all publicly accessible govt websites and data, etc etc.
The plugin presents a sidebar, or a popup in one corner, or even a separate tab or window. As you read an article/post on Facebook, Faux News... wherever... in real-time, the AI ingests each separate claim or statement, and flags them for veracity, with links for further research. So, in essence, a plugin that can call bullshit! in real time, with the relevant facts immediately accessible.
posted by Artful Codger at 9:54 AM on March 18, 2023
For the runup to, say, 2024, imagine an AI browser plugin, with training info from the ACLU, Southern Poverty Law Center, Snopes, wikipedia, universities, Library of Congress, all publicly accessible govt websites and data, etc etc.
The plugin presents a sidebar, or a popup in one corner, or even a separate tab or window. As you read an article/post on Facebook, Faux News... wherever... in real-time, the AI ingests each separate claim or statement, and flags them for veracity, with links for further research. So, in essence, a plugin that can call bullshit! in real time, with the relevant facts immediately accessible.
posted by Artful Codger at 9:54 AM on March 18, 2023
Eigenmorality: "anyone would be able to see, with the click of a mouse, the extent to which this parallel world had diverged from the real one."*
posted by kliuless at 4:26 AM on March 19, 2023
posted by kliuless at 4:26 AM on March 19, 2023
moral or amoral. […] It really doesn't matter what any of us think from our armchairs. It's a part of our reality now whether we like it or not
Eh? When something is becoming part of reality is exactly when we need to debate its morality and responses to it.
posted by clew at 9:56 AM on March 19, 2023 [1 favorite]
Eh? When something is becoming part of reality is exactly when we need to debate its morality and responses to it.
posted by clew at 9:56 AM on March 19, 2023 [1 favorite]
Uodate. Four days later and he has gained:
- 87,903 new followers on Twitter
- 1,501 on Discord
- $130 revenue
- 7,788.84 cash on hand (from investors)
• and has appeared on cnn.
Now 200+ other people have taken up the challenge to do the same thing.
posted by hoodrich at 10:47 AM on March 19, 2023 [1 favorite]
- 87,903 new followers on Twitter
- 1,501 on Discord
- $130 revenue
- 7,788.84 cash on hand (from investors)
• and has appeared on cnn.
Now 200+ other people have taken up the challenge to do the same thing.
posted by hoodrich at 10:47 AM on March 19, 2023 [1 favorite]
A decade ago someone basically undertook 1,500 slot machines walk into a bar (previously). It was basically an exercise in weaponizing Jenkins, but the gritty 2020s reboot would greatly benefit from today's AI. For the app name, ad copy, game design, code, fake reviews, logos, images, app submission, harrassing the reviewers until they approve, responding to bad reviews, etc.
posted by pwnguin at 11:04 AM on March 19, 2023
posted by pwnguin at 11:04 AM on March 19, 2023
Robo-trading is already a thing, just not something yet available to us mortals.
This $100 ChatGPT4 experiment is interesting. It may have the unintended consequence of making it clearer to the public just how easily our current iteration of capitalism can be gamed.
posted by Artful Codger at 12:13 PM on March 19, 2023
This $100 ChatGPT4 experiment is interesting. It may have the unintended consequence of making it clearer to the public just how easily our current iteration of capitalism can be gamed.
posted by Artful Codger at 12:13 PM on March 19, 2023
« Older Microsoft brings you the latest in eSports action! | Take the Vesuvius challenge: Help decode the... Newer »
This thread has been archived and is closed to new comments
this guy could have gotten an equally "viable" "business" from a google search. but, not being associated with the current fad, that would not have driven engagement on his twitter posts.
posted by your postings may, in fact, be signed at 10:24 AM on March 16, 2023 [29 favorites]