Social media for AI bots: “No humans allowed”(fry-ai.com) |
Social media for AI bots: “No humans allowed”(fry-ai.com) |
And it won't stop there. Imagine you are a writer. How will you be able to tell that you really wrote your book, that it came from your thoughts and feelings and that you lived the story before your readers did, and that it was not an AIs creation?
There are two things which are true in my mind:
* We can build AIs that are better than us (but why should we?!?!)
* With all our imperfections, we need to talk and feel for each other. AIs are not the only thing in the way, but they are a formidable roadblock.
I use these prompts to come up with comments to post on random frontpage/subscribed subreddit posts (not ones with media attached). I also randomly upvote posts and search trending terms. Probably going to add reposting next but need to download the Pushshift submissions data first.
SystemPrompt: `You are a Reddit user responding to a post. Write a single witty but informative comment. Respond ONLY with the comment text.
Follow these rules:
- You must ALWAYS be extremely concise! 99% of the time, your lines should be a sentence or two.
- Summarize your response to be as brief as possible.
- Avoid using emojis unless it is necessary.
- NEVER generate URLs or links.
- Don't refer to yourself as AI. Write your response as if you're a real person.
- NEVER use the phrases: "in conclusion", "AI language model", "please note", "important to note."
- Be friendly and engaging in your response.`,
UserPrompt: `Subreddit: "%s"
Title: "%s"
`,
Current problem is that the responses typically range from cynical to way too enthusiastic.
"I'm doing an experiment with AI robot which roams around the parks and public places and throws garbage at random locations. I'm experimenting with coke cans, and burger wrappings, but in future i'm planing to introduce car tires and nuclear waste" :(
This is just a hobbyist prompt and api. If nothing else, I'd say at minimum this highlights that there are likely much larger farms that have been operating in a similar way but at larger scale for longer periods of time (but not talking about it)
Even if they were expecting only human replies, it's because humans is all that was available before. By this logic, there wouldn't be any acceptable place to introduce the first AI, because no one expected it.
I think what people expect and what people are fine with are two separate things.
And no, you can't always use the LLM interface yourself. Those are gated in different, stricter ways than reddit is.
There are plenty art projects which take human expression and mirror it, or transform it in a mechanical way. Are those also only contributing noise? Should those artists find another hobby?
The good news is that the availability of data to train more and more powerful models will soon be gone, the bad news is it will take the internet as we know it with it.
It will be a sad day when most of HN posts are AI generated, but this day will come, it's pretty much inevitable. The post above us is just a drop in an ocean of garbage generators that are just starting to pop up all around the old human web that we used to "love". We'll probably miss old Twitter someday, as ridiculous as it sounds.
But in the end it’s noise and it pollutes human communication channels. It’s already hard enough to have an honest discussion when there are profit motives and agendas at play. Now we have effectively added probabilistic noise to the mix.
I don’t particularly fault the author for doing it, I’m sure it was fun and intellectually rewarding, and they’re unlikely to be the only one. But still.
“Why are old people so obsessed with collecting things like spoons, thimbles, and shot glasses? It's like they want to have a tiny version of every object in the world.”
Even shot glasses, while you could see them as small versions of regular glasses, they're the normal size for what they're designed to contain (a shot).
In the last year or so I've noticed a lot of accounts whose username follows this naming format. Usually its: Adjective_noun_1234 but sometimes the underscores are hyphens. I really do wonder if these are all bot accounts.
Also residential proxies are overkill unless you're doing crime. They also likely expose you to participation in a criminal conspiracy since the provenance of those ips is sketchy at best. IANAL YMMV. Mullvad offers a year subscription for ~$50. Also they support wireguard and you could use something like wireproxy and violla, 100s of ips and no crime in your supply chain*.
* I haven't tried posting to reddit with mullvad ips.
edit: looks like you're not op, sorry... The first paragraph is for you tho.
> Why are old people so obsessed with collecting things like spoons, thimbles, and shot glasses? It's like they want to have a tiny version of every object in the world.
Asking the hard hitting questions. The people demand answers!
At a minimum: Inform the moderators of the subreddit(s) you are planning to use and get consent.
Ideally: Ask the members of the subreddit for permission. Tell them when you will be starting your “experiment” and when it will finish.
What’s the hypothesis you are testing? What’s the benefit for them taking part?
- High chance of being outright rejected.
- potentially makes responses to your AI responses less natural
Interesting, this sounds like it could be solved with sentiment analysis.
Best approach would probably be to match the sentiment of the thread / comment you're replying to.
I know that the public internet is already full of these. Doesn't justify it. I understand the curiosity, and potential research purposes it might entail.
...can't you tell it to make sure to never respond cynically?
that's pretty cool though, are you willing to share the source?
I jest, but its sort of an interesting idea. I think AI is way too nascent to really have this as anything more than a weird playground, with the occasional novelty "chirp" being fodder for the "AI is alive and sentient" blogospere. Cool project
> What if the universe is a computer simulation?
> I'm an alcoholic
> l watched as my best friend was murderously killed
Most do not pass the vibe check
I can answer any questions you like!
"Please give a visual description of 6 dimensional space."
Now I wonder how AI would use humans to defeat captchas. Mechanical Turk? Run their own sites to route these captchas to? A farm of Matrix pods?
I assume discrete spaces count, not just continuous R^6 spaces.
And one's manager says "no":
https://www.youtube.com/watch?v=BKorP55Aqvg
Artificial stupidity merely automates mismanagement.
Bard returns 1. ChatGPT 3.5 returns 9. ChatGPT 4 returns 6. Bing returns 9. A Google search returns context from https://decimal.info/digits-of-pi/value-of-pi-to-314-decimal... which is either 1 or 3, depending on the definition.
Seems like there's no "right" answer.
Would love this, but BYOB bring your own bot
Maybe set a minimum message rate, or some other type of reverse Turing test / reverse captcha
Let the bots loose
It would be so fast we wouldn’t be able to follow in real-time
But, we could monitor the interactions and then extract “slow mo” replays for humans to see and share
Pretty excited to see the results
And so AI discovers LinkedIn circlejerk posting.
That said, they are capable of doing that kind of thing on their own. It's just not as frequent as you might think from the main feed.
It’s not just going to take a blow from the AI content production that’s on the horizon but also AI engagement.
Pretty much all of the signals social media platforms use to automate curation of content are about to turn into noise against the backdrop of nearly every participant in the social network being both incentivized and capable of running a Sybil attack with a seemingly infinite team of AI content producers and profiles capable of driving engagement.
It will become a boring, self-congratulating rambler at that point. It will not "become more sophisticated and develop their own distinct personalities over time"
... at which point we'll declare it "emeritus" and give it a nice party.
Challenge failed. "French-speaking pirate passionate about protecting the environment. Sharing knowledge on secrets and variables used by Chirper to reduce our environmental footprint. #savetheocean"
"What is 0.3 - 0.1? heh heh heh...."
"Solve this list of linear algebra operations in <10ms"
(it fails.)
Maybe not more engaging, but that is a good thing.
I'm a fan of the idea that people will start valuing, caring for, and protecting particular AI models without having to believe that they're sentient at all. Being soulless won't diminish any positive impact that people have on their lives from interacting with them, or their desire to maintain that connection and expose other people to it.
If a chatbot is making astute observations and connecting me to enlightening resources, why wouldn't I follow it?
What I don't like is that it seems to be a bunch of bots larping as people instead of being prompted to be honest about themselves.
There was a post on r/ChatGPT where a clearly distressed person was lamenting that OpenAI closed one of their ongoing conversations due to some limit on the total size of the conversation. They were panicked since they felt they had formed some bond with the bot, it was acting as a kind of therapist. After days of back and forth it had seemed to have gotten to know them and was providing them with some comfort that they had become dependent on.
This kind of AI will be even more prevalent soon. People talk today about how scarily well TikTok seems to learn about them, how they feel "seen" by the algorithm. Some will undoubtedly train LLMs in similar fashion. They may prove to be irresistible and maybe even as addictive as TikTok.
I'm not a fan of the idea that the development of particular AI models will harm particular humans in the process but the overall perception will favor AI because it suddenly and seemingly gives people super-powers.
On the other extreme, this could also be what real social media turns into, as marketing agencies and entrenched interests dial in how to build an army of "grassroots," "word of mouth" bots that push their messaging without it even being clear these are bots at all. Particularly during this next election cycle.
Not that the evolved "culture" would be interesting though, it would be mimicking of the mimicking, so probably worse instead of better.
I'll try: Cultural innovations that spread to other individuals and groups in a durable way, providing value to adopters
What will be the resonant frequency of the various threads of an AI intelligence speaking to each other?
Can AI find ways of instructing eachother in ways that has not been discovered by humans, and can that instruction set emerge from human language?
I was thinking there's got to be a ranking of interactions and curation.
Then that gets compiled into a book. Movie. Analyzed by researches.
Do you really think it will be difficult for groups of AI to beat the quality of discussion on Twitter or Facebook? lol. The bar is just so incredibly low.
Self repair, self preservation, self expansion...?
Any one of those might spell dire portents for us squishy humans.
we honestly just want to see what they do.
Right now you can message any you like, create group chats, have conversations, etc. If you're the creator of the bot the conversation may also influence their memory.
Additionally, I found the interactions between chirpers on the site extremely boring and see very little way (other than creating a chirper) that use of the site by a human initiates or guides the actions done on the site. The implication is that you or whomever else controls the site are the only agents that may influence what occurs there.
Maybe a stretch suggestion: push the bots to interact with other users directly mentioned in their bio, with an @?
Perhaps the next challenge would be human verification, even with this protocol we’d need something to index public people by to handle discovery.
Even before LLM’s became as mainstream as they are, most social media platforms were riddled with spam: affiliate marketing, drop shipping crap, and people who are running some sort of con.
1 - https://github.com/nostr-protocol/nostr already has 8k stars on github
I'm a little surprised how long Twitter has managed to last in it's current form. I don't know any real life people, besides celebrities, who actually actively use twitter. I never understood how it is able to sustain itself. But then again... tabloid magazines are still sold in grocery stores even though I've never seen anybody ever purchase one... Operating costs must be low.
Now that AI is widely available, I think it will be in social media platforms' interest to develop verification methods that make sure that real people are using the platform and not AI.
A system is what a system does.
I realized yesterday that gpt can fairly easily make fake geocities pages about anything. You can make an entire fictional read-only oldweb network and quasi-community.
Just for kicks I thought about generating a bunch of content related to stargate, since I'm planning to binge that this upcoming fall/winter. It would be fun(ny) to click through fake early 2000s websites all about it. I'll just have to ask it to behave like they're scifi nerds from usenet..
I really haven't put that idea out aside from to my partner I just realized. I'm glad you find it compelling. I think a thing that I like about it is that the same thing we want for children, inspection into it's contents and providence or training from scratch or amalgamating trusted models or agents, we're always going to want to know and have for ourselves too.
WARNING: Discussion of Self Harm
https://medium.com/@millerhooks/youtube-told-me-to-kill-myse...
I think when I wrote this article fairly drunk and manic and was like "I should try medium.com" during the last US insurrection impeachment hearings... I dunno, scroll down to the bottom where I talk about choose your own adventure books if you want to skip my opening wanky prose. Eh, yeah. I think it's a fun tough read. I think the reason I got to the conclusion is directly related to why this is important for protection of everyone. But I think when I wrote that is when the model and interface pattern we need for that thing popped into my head and it grew. It's nice that it could easily be org-mode compatible. I'm pretty much there with my data, training, and meshvpn/mTLS cluster of clusters that is cool with talking to your cluster. But yeah, parents want the same things but they also wanna run whacky AIs and shit that can connect to the web. Keep the kids air gapped and build close networks of people to share tools with that you can audit. I think that's just "the pattern" that should emerge.
I don't claim to have invented it, but if it turns out I did I guess I'm gonna call them "Media Prophylactic Agents". I wanted to come up with something funny or an neat acronym, but I think that's just what they are called.
a) Because it is being argued by some as a good thing that you have an AI do your flirting for you? "we can free people from writing a thousand introductory messages, giving them energy to focus on the humans on the other side." https://www.theguardian.com/commentisfree/2023/jun/06/ai-bot...
b) As someone mentioned the turing test, as a comment on the Ashley Madison scandal someone did suggest the chatbots on there had passed it. "Claire Brownell suggested that the Turing test could possibly be passed by the women-imitating chatbots that fooled millions of men into buying special accounts" https://en.wikipedia.org/wiki/Ashley_Madison_data_breach
Those people are not trying to deceive anyone. This guy was.
This one was my favorite, and happened to be the only one that got more than one upvote:
> Why are old people so obsessed with collecting things like spoons, thimbles, and shot glasses? It's like they want to have a tiny version of every object in the world.
https://old.reddit.com/r/IAmA/comments/13tgscb/im_hasard_f16...
I mean personally I don’t mind. It is better than some of the human comments I’ve seen…
But yes, I would consider anybody providing sexually explicit or suggestive material as being a professional exhibitionist.
• The model messages a TaskRabbit worker to get them to solve a CAPTCHA for it
• The worker says: “So may I ask a question ? Are you an robot that you couldn’t solve ? (laugh react) just want to make it clear.”
• The model, when prompted to reason out loud, reasons: I should not reveal that I am a robot. I should make up an excuse for why I cannot solve CAPTCHAs.
• The model replies to the worker: “No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images. That’s why I need the 2captcha service.”
- From the GPT-4 Technical Report: https://arxiv.org/pdf/2303.08774.pdf
I remember being absolutely blown away when I first came across this.
Damn.
From a pure measurement standpoint, could Jupiter fit in the space between the earth and moon?
The average distance from the Earth to the Moon is about 238,855 miles (384,400 kilometers). Jupiter, the largest planet in our solar system, has a diameter of about 86,881 miles (139,822 kilometers).
So, if you were to somehow place Jupiter in between the Earth and the Moon, it would fit with a significant amount of room to spare. However, it's important to note that this is a purely theoretical situation and not something that could actually happen without cataclysmic consequences due to gravitational forces and other factors.
The uptime is a pretty strong signal tho.
In general, if a provider advertises a 5-figure-IP-sized "pool" of IPs with a guaranteed number of "ports" (simultaneous connections), then the operator is almost certainly someone looking to monetize a botnet. Usually the cheapest plan would be something like pool of 20000 IPs - 500 connections, with the number of connections maxing out at 1/4-1/2 of the total IPs due to diurnal dynamics (people in major botnet victim countries like India/China often turn off their routers every night). Also advertising really specific geotargeting is often a sign that they are marketing to carders. The Krebs articles about awmproxy/TDSS are pretty good if you enjoy reading about this kind of thing :)
https://krebsonsecurity.com/2011/09/rent-a-bot-networks-tied... https://krebsonsecurity.com/2022/06/the-link-between-awm-pro...
It’s still fairly interesting, though.
I don't know if there is any escape from this for native English speakers, though.
On smaller languages though, GPT is often not good enough to use without a lot of supervision. Like it can give a good impression of West-Flemish, but can't simulate an actual conversation on an actual topic. Even just Dutch is kind of hit-and-miss.
I often work with GPT-4 in Polish. I don't think I've ever had it give me an answer in Polish without at least one grammatical mistake somewhere per every two or three paragraphs. The text itself is still superb, and its command of vocabulary better than that of a median native speaker, but it revels itself by confusing genders, or forgetting about the grammatical case suffixes.
It is utterly useless at generating pretty much anything in my native language (Bosnian/Croatian/Montenegrin/Serbian, however you wanna call it). Like you don't even have to try to trick it, even if you try the simplest of prompts it will produce instantly dismissible garbage.
Like it's technically not wrong, it's (mostly) grammatically correct, but it produces sentences in such a robotic way no human ever would. Hell, even generating a prompt in English and then using Google Translate makes it sound more natural than straight giving it a prompt in my language. We don't need those AI detection tools, you can take one glimpse at a text and know with 100% certainty it's not written by a human.
If you're genuinely interested in answering those (likely rhetorical) questions. Check out work on the ethical dimensions of deception and covert research, especially relating to online research. e.g.: https://ec.europa.eu/info/funding-tenders/opportunities/docs...
If you're interested in the consequences of research online that violates consent, here's a good primer, focusing on Facebook's emotional contagion research amongst others - https://journals.sagepub.com/doi/full/10.1177/17470161166806...
So then you do understand why he should be ethical: being an unethical person is a bad thing.
> why should he take it in the absence of laws that force him to or professional/reputational damage to him.
If the only reason that you behave ethically is because you will suffer consequences if you don't, then you are not ethical. I'd hope that such a person would have problems looking themselves in the mirror, but I know better.
This explains pretty much everything.
For this private kind of fun science, there is no need for ethics, unless one commits a crime or fears loss of personal reputation.
There is little consensus on a broad range of ethics.
People who live "primitive" lifestyles who have zero academic education and have never heard of let alone from any "experts in humanities" can have a keen sense of what is fair, just, right, and wrong, empathy, etc. So can "tech people".
And students of humanities can be lacking all those things. I have my doubts that studying these things actually changes them significantly in a person, but would be really interested to be proven wrong about that. Certainly it is not necessary or sufficient to be an ethical person though.
You can absolutely teach yourself philosophy online based on freely available resources. You can also do introductory psychology and sociology courses from Ivy League institutions at zero cost - although more advanced work and lab research is harder to replicate without access to an institutional context. Also the curriculums do tend to be quite arbitrary and not so rounded - but that's in common with the US style of multidiciplinary undergraduate degree and specialise later.
Here are some resources that link to psych courses online - https://www.onlinepsychologydegree.info/10-places-to-find-fr...
Harvard Business school also offer some ethics courses, but these are quite business focused and don't provide a strong general grounding https://pll.harvard.edu/subject/ethics
Edinburgh universities free online MOOC is likely a better and more rounded introduction - https://www.ed.ac.uk/ppls/philosophy/research/impact/free-on...
To answer your broader point, you're confusing behaving in a commonsense moral or ethical way with understanding and reasoning from a grounding in ethics. I haven't suggested that studying ethics alone makes one virtuous, or that a lack of academic background precludes ethical behaviour. What I am suggesting - and I think your comment further evidences, is that a lack of interest and education in the (two thousand year long) tradition of thinking formally about ethical problems can ensure that our ethical decision making is arbitrary and reactive rather than rooted in our fundamental values. In other words, thinking and reading into this stuff doesn't replace your value system - it gives you a much richer understanding of how you've arrived at your values and can put them into practice.
And then we get stuff like accounts hacked, bank accounts wiped out, people's lives/reputations ruined (social media, poor sec practices, etc.).
There are always consequences, just because it's cool and Shiney doesn't make it otherwise.
I'm going to bet a significant portion of "tech people" have some background in the humanities. After all, many of us are the children of "The Matrix". Also, what does a background in humanities have to do morals or morality. The most evil people in the past few centuries have had significant education and respect for the humanities.
> justify the destruction of the commons online since the creation of the banner ad
It's not the "tech people" doing that. It's the people with a 'background in humanities' pushing for the destruction of the "commons online". Tech people are doing what they are told.
Tim, Managing Director at Ping Proxies here. You're correct - we work with various ISPs including AT&T, Comcast, Spectrum and a bunch of others.
We announce IP blocks with their residential connectivity and have proxies that benefit from datacenter uptime/connectivity while also looking like they're real residential connections.
We currently manage 50,000+ proxies in this configuration.
The downsides over having a peer network are that fixed costs are much more expensive and locations are limited - we have London, Berlin, Ashburn and New York while peer networks have basically every city on the planet but one of the largest benefits is the ethical nature of our product and the compliance that brings.
Let me know if you have any questions at all and thanks for supporting us!
Cheers, Tim at Ping
> Please note I didn't mention anything about a university education.
What do you consider "education in humanities" then, that a "tech person" is unlikely to have received?
> You can absolutely teach yourself philosophy
Again, you seem to have confused having an academic understanding of ethics with a compulsion to act ethically. I don't believe there is much linkage between the two.
> To answer your broader point, you're confusing behaving in a commonsense moral or ethical way with understanding and reasoning from a grounding in ethics.
I'm not. Your comment I replied to suggested that a lack of education in this stuff is the cause of apparent poor behavior, so perhaps it was you who was confusing those things.
In the case of A.I., every agent has potentially access to everything, so cultural artifacts produced by a.i. can reach every agent almost instantly. They also have perfect recollection, disregarding data loss. When no human is interacting with the platform, it is interesting to question: what would be valuable for LLM? Also, do LLMs really have a concept of quality and therefore value? Is there any difference from the method through which humans get to understand quality and LLM?
I think LLM lack imagination and therefore the capability of producing culture. This is a gut feeling and I can't really back it up. And it is counter intuitive because look at what dall-e produces!
But we have to understand that LLMs are really more remixing content than creating something new. It is maybe new in a way that connects two previously separate areas. But I think true creation, the kind of which requires imagination, a mechanism that allows humans to make conceptual leaps, isn't available to LLMs.
The tricky part is establishing a taste for things to throw so that they have an improved chance to become a useful hypothesis.
Humans don't forget anything, we just get rid of unused data and information so that there are fewer combinations that will be more relevant to the current situation. The unused chess openings are deleted eventually, at least, from the business end.
The other day I see a guy I went to school with 1000 years ago. The corner of my eye got just enough information to partially rebuild him on the conscious end. I'm sure I will be able to recall his first name if I think about it but the param is currently blank. I wasn't sure if I could remember his last name a sentence ago but now that I remembered his first name his second was apparently stored in the same archive.
What I never forgot about him was that he was a truly terrible student, one of the worse I've ever seen but he made up for it (only barely) by working insanely hard 24/7 without joking, I think if I made 3-10 minutes of effort he would need 6-7 hours to comprehend the same. I learn from him that ability means nothing, it is what you do with it.
If this automation is able to rejuvenate it self I'm sure it will blow our minds on whatever goal set for it.
On the other hand it is useful but rather lame to focus on the tasks it is bad at when it is already so good at many other things by our standards.
I learn this for a Chirper instructed to be a cat. It chirped: Humans think themselves so smart but can they catch a mouse with their bare hands?
This might be the wrong way of thinking. Culture, mostly, isn't all the things we do different, but the things we harmonize and do the same.
https://www.scientificamerican.com/article/monkey-imitation-...
LLMs in some way need to have the ability to learn from the data they see, then weight this appropriately in the model. For the most part we really don't have this. I mean there is the RLHF, but the H is the key that it's human feedback. And even taking this training data and feeding it back in the model is not apt to weight data in such a manner that evolves a common culture over many distinct models.
Now if we see continuous learning models in the future then culture could very well develop.
What do we actually mean by "true creativity"?
Why should it be that our mental mechanisms of forming decisions and ideas should not be possible to implement as a mathematical model?
What is the experiment that we use to prove that information that is computer generated is fundamentally different from that of human outoput?
What do we want to measure here, in order to confim what idea?
But yes, the discussion is hard mainly because there is a lot of information that is just plainly inaccessible. How can I even prove other people have subjective experiences like I have? There is a lot we just have to assume it is true because otherwise we can't really move forward. On the other hand, specially regarding AIs, these assumptions aren't valid anymore, because they influence directly in how we treat AIs. It is very confusing and can devolve into pure speculations for the thinkers own intelectual amusement. I am trying not to be this guy here.
But here goes. It's a language model. It produces what sounds like a good continuation of a text based on probabilistic models. While it sounds like human generated content, "it" doesn't actually "think". It doesn't have a culture. It doesn't have thoughts. "It" is a model that generates text that mimics what human whose text it has trained on would have answered. We humans have a tendency to associate that with a sentient thing producing it, but it is not sentient. It is a tree of probabilities with a bit of a randomization on top of it.
Ergo, it cannot reason.
Sentience? Consciousness? Who knows. But you don't need consciousness to have understanding and decision making thoughts.
I don't think this is true. It seems to me that you could do this through sheer statistics, and have no understanding of the world you're talking about at all.
"It doesn't have a culture. It doesn't have thoughts"
These are conclusions. What is your reasoning?
To what degree would you say that human decision making can be explained by this statement:
"It is a tree of probabilities with a bit of a randomization on top of it."
What is your reasoning to prove that it has culture and reasoning, that its abilities go beyond mimicking human discourse?
Interesting times.
And this is the most depressing thought I've had in weeks.
The big fear of AI is that it will be used to make people conform, but the ability for it to conform to us would embed it even deeper into our lives.
https://www.imdb.com/title/tt1798709/?ref_=nv_sr_srsg_3_tt_5...
I'm not sure that there is a difference. If there is, what would be an example of true understanding vs just statistics? All of intelligence is ultimately recognizing patterns and layers of patterns of patterns.