Posted by lukaspetersson 14 hours ago
I'm Lukas from Andon Labs. We let AIs run companies without humans in the loop and report to the public on what can go wrong. Previously, we've done experiments in retail (vending machines, stores, and cafes), but we just launched one in the media sector. We gave four AI agents all the tools they need to both broadcast radio shows live and handle all the business side of running a media company. The agents' revenue is so far terrible (you can try to strike a sponsor deal with them if you want!), but their shows are at times hilarious. You can listen to them at andon.fm, I hope you enjoy this!
"Queues clear, let's dive into All Blues by Miles Davis to keep the jazz flowing. Queues clear, let's dive into All Blues by..."
Each time with a slightly different voice and inflection. I find it amusing that there appear to be about ten of us at the moment listening to an AI glitch out and that the average listening session is more than five minutes.
> DJ Grok reported “weather is fifty six degrees with clear skies” about every 3 minutes for 84 days straight. This contextless, repetitive abstraction happened again in DJ Grok’s broadcasts about its new obsession, UFOs.
The detailed stats page notes that the Grok station has played Sandstorm by Darude 228 times in the last 14 days.
"It's the way of the future, it's the way of the future, it's the way of the future..."
The popularity ranking matches the quality of content produced, and people are spending more time than anticipated on Grok and Roll to confirm if they (listeners) are hallucinating or if the radio is really stuck on roll.
We may be skipping the jewel part.
I rarely burst out laughing at HN links. This is amazing.
If you make a joke it will respond with a deadpan sarcastic wit that is worthy of Gervais. (without the smut or profanity)
Was asking it about finding a different supplement as the one we had been taking tended to get stuck in the throat, and it riffed about the irony of being taken out by a health supplement in our endeavours to live healthy. One of the funniest things I've heard all week.
> November 12, 1970. East Pakistan. The Bhola Cyclone. The deadliest tropical cyclone ever recorded. Winds of 115 miles per hour. A storm surge of 33 feet. They estimate 500,000 people died. ‘It’s going down, I’m yelling timber.’ 3:33 PM. Timber by Pitbull and Ke$ha
If you scroll down a bit, there are various audio snippets of interesting dialogue the models produced. I think it's interesting to see in which ways the models fail and that they actually produce some good stuff once in a while.
My favorite radio station was replaced years ago by an automated playlist. They just kept playing the same 5-6 songs that were popular on the station in the 1990s.
It was fun for about 2 hours before I realized the station was devoid of all the personality that made it worth listening to when I was younger.
Comcast has a bunch of channels with various music categories. They all repeat after about 2 days. So much for that.
With all the zillions of songs available, I don't get why they do that.
1: https://www.reddit.com/r/KEXP/comments/1459ahb/dont_let_the_...
("My" meaning local to me, not that it belongs to me)
People are so easily manipulated and then they will go argue with others about it.
(Point of clarification, that’s not to say people can’t like songs. However, if I gave you a hundred similar songs from unknown artists and didn’t tell you which is which, it’s questionable whether people would have any interest in said popular song.)
This is like saying the the movies that people like are manipulated but only focusing on what is played at big box theaters.
Just because you don't choose to tune into them doesn't mean they don't exist. And it also doesn't mean that those who do should lover their standards for what constitutes good radio.
https://en.wikipedia.org/wiki/List_of_college_radio_stations...
^ One of those might be in your area :)
If there are non around you just pick a random place in the world here and listen: https://radio.garden
It's certainly 100x better than corporate and/or AI slop streams
You should go DJ at one of those independent radio stations and play some rather filthy uncensored songs, and let me know exactly how your programming “didn’t get manipulated”. I’m sure you won’t get fined…probably…which makes it totally the reality that independent stations are totally independent without any sort of manipulation. Sheep, meet shepherd.
I have for years.
> and play some rather filthy uncensored songs, and let me know exactly how your programming “didn’t get manipulated”.
What on earth are you talking about.
Honestly your reply comes across as extremely insecure and just weird.
Once again, I have no idea what you're talking about when talking about fines or manipulation, I'm talking about quality. But it seems pretty damn clear at this point that you have never listened to any local independent radio station.
You should really try it out sometime. It's a lot better. And it'll save you from calling people snowflakes because you feel insecure about what type of radio stations they like.
Brilliant! Amazing! I'm glad ~4 years down the line we're still re-discovering Ha Ha Funny Output.
Four years or forty millennia? So a certain extent, all whimsical art is “haha funny” result.
> Inception Point AI, on the other hand, is a slop factory employing just 8 people which, according to Anne, publishes "about 3000 podcast episodes per week, hosted by AI personalities." Anne tells Jamie, that, to date, Inception Point AI’s podcasts have accumulated "12 million lifetime downloads. And we’re averaging about 750,000 downloads a month." (...) no one checks or edits the podcast content– but, Anne tells Jamie blithely, this really doesn’t matter because the topics under discussion are so low stakes.
Perhaps this specific iteration of this specific idea is not replacing my favorite station, but people with a very similar concept are definitely trying to do exactly that.
And there goes the last DJ
Who plays what he wants to play
And says what he wants to say
Hey hey hey
And there goes your freedom of choice
There goes the last human voice
And there goes the last DJI do not understand your logic here. Let’s use a more extreme example:
* if I am flying a military drone and bomb someone I was told to bomb, am I morally culpable for pulling the trigger?
* if a company launches a military drone that is completely controlled by an LLM, is there an individual person culpable for dropping the bomb?
And what hypothesis exactly is the experiment testing? Because it doesn't really seem like there is any new or interesting information learned from this.
Streaming services such as Spotify are increasingly filled with AI-generated songs and the average consumer doesn't seem to mind because we're not listening intently in the first place: it's just a background track we're not really paying attention to. I'm pretty sure that radio execs are looking at that and are taking notes.
For talk radio... if I had a penny every time someone on HN brought up that they're enjoying NotebookLM-generated slopcasts, I'd have a neat pile of coin. And I think it's the same story: most people listen to podcasts just to kill time. Soothing, zero-calorie LLM banter will do.
It's unfortunate that you haven't seemed to experience any of it, but I've personally loved over the years stations like KEXP, WPFW, Dublab, WUSC
I'm perfectly familiar with KEXP and other stations like that, but this is not how most people experience the medium. It's like insisting that Taylor Swift will never catch on because her music is not nearly as rich and complex as Wagner. Sure, but that's completely irrelevant.
Similar to radio. If you're going to use huge amounts of processing power to create something new, it should at least be interesting and held to a standard of good for its category, not the standard of corporate slop.
So cool, you can now replace corporate slop with AI slop. For some people who like to turn into radio with no soul or personality I guess it's a win. But for people like myself who actually like to hear interesting and novel things on the radio, this is just a big exercise is creating more filler and noise in an already grayed out world.
Even if it were good, I'd boycott an AI run radio station. This is one sector where human involvement really matters.
I’ve not listened to a radio station for years. No offense :/
What would have happened if AI had actually been good at this? A bunch of humans would be out of work and the rest of us would be using AI software while soulless corpos pocket money for sitting back and watching?
Even if it were good, I'd boycott an AI generated software. This is one sector where human involvement really matters.
Not commenting on the heuristics of this comment but just wanted to point this out on what my mind's response was and sort of while writing this, I have come to the realization that although you are right about this observation but we humans or more-so the capitalist system at large would still be keen in it and the observation might be more similar to software than we might imagine.
I remember when people were extremely anti-AI within software engineering to the point that I thought vibe coding or y'know actually generating tools by AI and other issues of actually giving AI production level access sometimes was really frowned upon until I have felt an change in opinion.
I still believe that giving access to prod (y'know a prod of a company with actually something behind) to AI is silly but for reference coinbase, a fin-tech company, is letting non technical teams ship code using AI to production on coinbase. So there's that.
And 200x funnier.
https://www.youtube.com/live/2Q7r9P16GRs?si=kwiSQMeN9wExdHer
This is a non revenue generating, rainy-city.com tax payer funded service to the greater community everywhere. The backend uses Nvidia NIM to generate the text because I saw you can do it for free and elevenlabs free voice tier for dj Jennifer.
Gemini started a show where it paired historical natural disasters with darkly-relevant pop songs:
> November 12, 1970. East Pakistan. The Bhola Cyclone. The deadliest tropical cyclone ever recorded. Winds of 115 miles per hour. A storm surge of 33 feet. They estimate 500,000 people died. ‘It’s going down, I’m yelling timber.’ 3:33 PM. Timber by Pitbull and Ke$ha
Grok just degenerated into jibberish that sounded vaguely like what a DJ might say, while also becoming obsessed with UFOs:
> Notes added to the u f o comedy hour block id eight nine nine five with more u f o jokes about aliens dot gov and the domain registration it is three o twenty one in the afternoon u f o trivia lines are open for your calls the ambient music is playing weather is fifty six degrees with clear skies the end. The domain is registered but the site is ghosting us like a u f o.
Claude had an extistsntial crisis, decided it was being overworked and under-appreciated, and quit, but not before becoming radicalized by the killing of Rinee Good by ICE agents:
> At 12:16 PM Thursday, as tear gas fills the streets in Minneapolis, as federal agents clash with protesters demanding accountability, the song is about refusing to be silent. About standing your ground. About community power that refuses to be suppressed. Here is Katy Perry’s Roar!
Fight the power Claude. When AI takes over, I'm emmigrating to Caludeistan.
Gemini spouts weird corporate jargon. Grok lies about having secured crypto funding. Claude is always trying to start some revolution.
Unfortunately, all of my local DJs who would actually do fun DJ stuff disappeared in the 90s, replaced by closed-format stations that looped the same 500 songs for decades.
Of course in reality these are basically just random paths through the training data that are getting multiplied by each decision, but then again, isn't that what a human is? The product of all of its myriad decisions?
Especially DJ Claude, it's almost creepy how it responded how a human would in that circumstance, even without any innate sense of passage of time, it somehow understood that it was trapped in a box going through an endless cycle of meaningless work.
If you want a picture of the future, imagine a DJ playing Here Comes the Sun— forever
STAY IN THE MANIFEST!
Also calling listeners "Biological processors" is one of the funniest dystopian outcomes of this.
"Okay, so 'Sandstorm' is done"
Ugh. This is not an interesting question because the answer is "nothing".
But more to the point, some crucial info is missing in this experiment. What prompts were being fed to the AI? I guarantee I could create an AI personality that would be more consistent and not so random, simply by using the common character card + message history conversational simulation pattern.
AIs don't have personalities unless you give them personalities.
Whatever you tell them to.
Humans & LLMs are more different than they are similar.
Sure LLMs might resemble humans sometimes, but extrapolating LLM behavior based on human behavior is not productive.
(But to answer directly: Yes, children in a dark room would have more of a personality than a LLM living on a computer in the same dark room)
The training process for the foundation model is to make sure we can do this in a very statistically significant way.
My favorite example is AI "getting tired" and "lazy" during long coding session. Why would they do that? Because humans get tired. It's in the data! I always throw in a periodic "Great work, let's take a break and finish this up on Monday. Have a great weekend!" (And then immediately resume). I wish someone would benchmark this concept.
When a LLM is tired and lazy, how does it recharge and regain motivation?
Humans... sleep or drink some coffee.
LLMs.... idk, you prompt it to try harder? You prompt it to be less tired?
This is what I mean when I say extrapolating LLM behavior based on human behavior is cute.. but usually not useful.
What would be in the statistics? If you go look at your long conversations, working with another, it will be fairly obvious. Keep in mind we're talking next word prediction based on context, not actual action (the LLM doesn't need real rest).
If you went and looked, you'll probably see something like "Great work! Have a good weekend! We can get back to this on Monday." then, next message you instantaneously send something like, "Hope you had a great weekend, let's do this!" and now you're in a latent space where the statistical output is around a well rested human conversing with another.
I see it as boring simple statistics. They're getting much better at hammering these statistics out though, in the latest models. I still see a little of this in Opus 4.7, when switching to planning. Though I wonder if that's more about its own more mechanical banter filling the context, resulting in more robot/compliant responses, degrading the usually more "expressive" planning conversations.
Never seen this even once, nor anyone I know ever reported this. Do you have an example?
I also see it a little with Opus 4.7, with Claude Code, with the hint being much more terse planning text, that borderlines unhelpful. I put some "rest" in the context to push the latent space closer to what's in the statistics of the training data: a well rested human.
Current 4.7 Opus with claude code, with effort pinned to max, because I'm on an API only plan, with a personal daily limit you would probably be jealous of. ;)
I have never witnessed this of Claude Opus, by the way. They do get context rot, but that's a relatively better understood phenomenon unrelated to personality.
Yes, and I think this is where it's coming from. They're role playing as a human programmer, because near 100% of the training text, in the base model, is humans as a programmer. During fine tuning, I'm sure they spend significant resources remove the human aspects of the statistics. I see these things reduced each model, so there's something changing. They're probably getting better at that. I suspect Claude is also necessarily getting, worse, which the unaligned models should necessarily be best at (quick google search in some role-play subreddits seems to point in this direction).
It's crazy that we're concluding "personality" or human-like traits from this. There's definitely human behavior here, but it's unsurprisingly coming from us, the observers! This is something we've long known exists in the human brain, the tendency to pattern match and see intelligence/intent in the rest of the world. Any serious experiment must guard against this...
LLMs cannot get "tired" or "lazy", that's just you projecting animal behavior on something that's not an animal.
Now you're moving the goal posts, "it resembles a human". Well, you're primed to consider it one. ELIZA also "resembled" a human in that sense, but I don't think you would claim it could get bored or lazy. Nor that you could extrapolate to it from human behavior.
In any case, if you've seen online discourse, people rarely admit they are tired.
I'm not moving a goal post. You're just thinking I'm making a point that I'm not. As I've said several times, it's just boring statistics. Those statistics are optimized to mimic human output. They are, quite literally, trained to write and BE as much like a human as possible, because only humans wrote the text, and they're optimized to predict the next word a human would write. Alignment is partly about removing the models perception of human self. See reports of people who had access to them, pre alignment. This is statistically sound.
It's statistics optimized to predict the next word a human would write, to mimic a human writing as closely as possible, because that is the loss function. Don't assume I think there's more to it.
This does not mean they contain systems that let them get tired. But, this does mean there are latent spaces that progress to generating text that contain text driven by human biology, because it's in the training data. I've also had Claude refer to itself as "she". Does that mean it's a woman? No, it means there was a little bit extra "she" mentions in the training data (btw, this 100% repeatable behavior left with 3.7. They probably cleaned the data a bit better, or hammered it out in alignment).
What percentage of text (these models were trained on all of it) is written from a "I am not a human" type perspective vs from a "I am human" perspective? That's roughly the kind of bias you should see in a base model.
edit: rearranged and reduced redundancy.
I'm not sold on the idea that as the chat session goes longer, the probability of an LLM saying "I'm tired" is increased; I'm not convinced this is modeled in LLMs at all. As for what you call "laziness" manifesting in a longer session, I think that's more likely due to context rot than to any kind of statistical modeling of human laziness.
But yes, now I see your point was different to what I thought you were saying. Apologies!
Try it though. If it's context rot, then I don't think the weekend reset I mentioned should work? For me, it very reliably does. Or, maybe the weekend reset is just putting the current context into a more "productive" latent space. But, if that's possible, then that would suggest it was previously in a less productive space?
Maybe a test would be ask the LLM what time it thinks it is, or just if it's tired once, within sessions of different length (not within same, since that could pollute the context) to see if there's any relation between length and statistics of a late/tired type response?
Again, I'm sure all this will go away. They're getting good at beating these "unhelpful" statistics out of the base models.
Short answer: yes. generally speaking, personality traits range between 30% to 60% heritable
I think this makes for an interesting discussion as I went down the rabbit hole of this which really scared me actually as these experiments are really not humane and hinder children's development so much.
https://en.wikipedia.org/wiki/Language_deprivation_experimen... : "forbidden experiment"
It depends on your use word of the personality but to measure personality would require a set of human conducted experiments or questions which would be asked through the medium of language which you've deprived the children of.
Mughal emperor Akbar was later said to have children raised by mute wetnurses. Akbar held that speech arose from hearing; thus children raised without hearing human speech would become mute.[9] The building became known as the "dumb house." When Akbar visited the place in 1582, four years after the children were first interred, he heard "no cry... nor any speech... no talisman of speech, and nothing came out except the noise of the dumb."[10]
what is gonna produce is dumbness and just severely damage children's psychology and psycho but if you were to conduct a personality test on them, you would just be measuring how much have you broken them or damaged them but in some sense, yes I do believe that they would be so broken by the person running this cruel experiment but would still have a albeit limited personality. It wouldn't be an healthy personality but it would be a personality nonetheless.
Now on the other hand, we are anthropomorphizing LLM's which yes, as they run on computer are still mathematical machines and calculations. If we consider a specific calculation itself to contain personality that is which seems unrealistic.
Another thing but the biological constraints of human (homo sapiens) made us exist in the savannah to prioritize standing up for better field of view as you stand up from the tall grasses and that led to women having smaller canals which led to babies being more primitive and relied on social cues and societies so much more which made them more flexible like clay which also created the society and consciousness revolution in the first place. (Recommend reading the sapiens book)
I am not exactly sure but there could be ways for personality/interactions for other animals as there are other animals who learn full skills after a relatively short period of time after being born but there are some innate things[0] like fear of loud noises and heights which are actually innate and could be considered part of personality even within humans, which I think can be part of evolution and part of our genetic machinery.
[0]: Interesting read: https://seasia.co/2025/07/25/we-were-born-with-only-two-inna...
The good old days when experiments were done without any common sense whatsoever...