Pinned Tweet
I made Plague Inc. but for ideas.
You can sow any idea in a simulated persona's mind and watch how it spreads through the city
The goal: get as much of the city to adopt and believe in your idea as possible
Compete with your friends for the top spot on the leaderboard. Link ↓
You can sow any idea in a simulated persona's mind and watch how it spreads through the city
The goal: get as much of the city to adopt and believe in your idea as possible
Compete with your friends for the top spot on the leaderboard. Link ↓
2
0
6
439
Egor retweeted
Honoured to be featured by @JacobReid02 at @Bloomberg.
I think what we're building has great potential for better policymaking: no more guessing what would work, or worse, risking real societies on "bold policies" with unknown consequences. Now, policymakers can test what will happen in artificial societies, before taking the bet that impact all of us.
At the same time, I strongly believe that what we're building *must not* replace talking to real people. Because there's more than knowing what policies will work in a simulation - policymakers should talk to real people, build real empathy, feel real people's real joy and pain, so that they can be effective representatives in our democracy.
Simulation must be used to better serve the people, not to replace their voice.
After all, that's why we exist as a company: to unfold humanness, and help scale humanity’s ability to collaborate and innovate effectively.
I think what we're building has great potential for better policymaking: no more guessing what would work, or worse, risking real societies on "bold policies" with unknown consequences. Now, policymakers can test what will happen in artificial societies, before taking the bet that impact all of us.
At the same time, I strongly believe that what we're building *must not* replace talking to real people. Because there's more than knowing what policies will work in a simulation - policymakers should talk to real people, build real empathy, feel real people's real joy and pain, so that they can be effective representatives in our democracy.
Simulation must be used to better serve the people, not to replace their voice.
After all, that's why we exist as a company: to unfold humanness, and help scale humanity’s ability to collaborate and innovate effectively.
Politicians including @KanishkaNarayan and @darrenpjones are testing ideas on AI-generated voters from companies like @societies_ai. Useful for targeting comms, honing arguments, training canvassers - but humans are weird, and tough to replicate!
bloomberg.com/news/articles/…
1
1
15
587
Egor retweeted
Increadibly exicted to announce Rothko to the world! For me, it’s been a labour of love that started three years ago when I co-authored the first large-scale paper on AI swarms with @james_k_he.
During this research, James and I noticed that when language models interact at scale, they mimic fundamental characteristics of human collective behaviour without being explicitly instructed to do so.
These findings suggested something exciting: pre-training bakes an inherent understanding of human group dynamics into language models.
Despite our excitement, we also noticed that RLHF works against LLMs’ emergent simulation capabilities, compressing the nuances we observe in human behaviour and pulling models towards a population mode.
Hence, we faced an interesting question: can we use and enhance language models’ pretraining knowledge to accurately simulate human societies?
Rothko represents the culmination of these efforts. It works like a decision model trained on signals we’ve extracted from a large corpus of human behavioural observations, including anonymised social media information, transaction histories, unstructured interviews, and more.
By getting rid of RLHF post-training, moving from the domain of natural language to the domain of decisions, and training on behavioural data, we’ve seen that Rothko dramatically outperforms SOTA LLMs at accurately simulating individuals and groups.
This finding underpins our work at AS and the value our simulation platform provides to clients.
Looking ahead, I see a world where large decision models like Rothko play a fundamental role in scaling innovation, protecting sovereignty, and aligning AI swarms with human values.
I find these applications incredibly exciting, and I look forward to scaling Rothko’s capabilities with the awesome team at AS!
During this research, James and I noticed that when language models interact at scale, they mimic fundamental characteristics of human collective behaviour without being explicitly instructed to do so.
These findings suggested something exciting: pre-training bakes an inherent understanding of human group dynamics into language models.
Despite our excitement, we also noticed that RLHF works against LLMs’ emergent simulation capabilities, compressing the nuances we observe in human behaviour and pulling models towards a population mode.
Hence, we faced an interesting question: can we use and enhance language models’ pretraining knowledge to accurately simulate human societies?
Rothko represents the culmination of these efforts. It works like a decision model trained on signals we’ve extracted from a large corpus of human behavioural observations, including anonymised social media information, transaction histories, unstructured interviews, and more.
By getting rid of RLHF post-training, moving from the domain of natural language to the domain of decisions, and training on behavioural data, we’ve seen that Rothko dramatically outperforms SOTA LLMs at accurately simulating individuals and groups.
This finding underpins our work at AS and the value our simulation platform provides to clients.
Looking ahead, I see a world where large decision models like Rothko play a fundamental role in scaling innovation, protecting sovereignty, and aligning AI swarms with human values.
I find these applications incredibly exciting, and I look forward to scaling Rothko’s capabilities with the awesome team at AS!
Introducing Rothko, a foundation model of human behaviour designed from the ground up to simulate the world’s most consequential decisions.
Rothko is the result of over three years of frontier research at the intersection of behavioural science and machine learning.
We’re only r the model’s value across domains ranging from market research to AI safety and swarm alignment.
1
2
12
587
Egor retweeted
Introducing Rothko, a foundation model of human behaviour designed from the ground up to simulate the world’s most consequential decisions.
Rothko is the result of over three years of frontier research at the intersection of behavioural science and machine learning.
We’re only r the model’s value across domains ranging from market research to AI safety and swarm alignment.
Rothko is the result of over three years of frontier research at the intersection of behavioural science and machine learning.
We’re only r the model’s value across domains ranging from market research to AI safety and swarm alignment.
23
22
396
20,894
Too many people were complaining that the guardrails were too strict.
So I loosened them up a bit. Go wild. See how far your idea travels.
So I loosened them up a bit. Go wild. See how far your idea travels.
Can your idea become the talk of the town?
We modelled how thoughts and ideas spread in human societies by putting 2,721 simulated personas representing the US general population in an artificial city. Each persona has its own schedule, beliefs, and social circle.
Sow an idea watch it take hold and go viral, or see it fizzle out if no one takes interest in it.
Compete with your friends and colleagues to see whose idea can win over the most adopters, get a high score and put your name on the leaderboard. Link below.
0
0
5
210
Egor retweeted
This is a social scientist's wet dream.
Can your idea become the talk of the town?
We modelled how thoughts and ideas spread in human societies by putting 2,721 simulated personas representing the US general population in an artificial city. Each persona has its own schedule, beliefs, and social circle.
Sow an idea watch it take hold and go viral, or see it fizzle out if no one takes interest in it.
Compete with your friends and colleagues to see whose idea can win over the most adopters, get a high score and put your name on the leaderboard. Link below.
14
92
1,569
117,871
Egor retweeted
Can your idea become the talk of the town?
We modelled how thoughts and ideas spread in human societies by putting 2,721 simulated personas representing the US general population in an artificial city. Each persona has its own schedule, beliefs, and social circle.
Sow an idea watch it take hold and go viral, or see it fizzle out if no one takes interest in it.
Compete with your friends and colleagues to see whose idea can win over the most adopters, get a high score and put your name on the leaderboard. Link below.
We modelled how thoughts and ideas spread in human societies by putting 2,721 simulated personas representing the US general population in an artificial city. Each persona has its own schedule, beliefs, and social circle.
Sow an idea watch it take hold and go viral, or see it fizzle out if no one takes interest in it.
Compete with your friends and colleagues to see whose idea can win over the most adopters, get a high score and put your name on the leaderboard. Link below.
72
186
2,947
426,476
Egor retweeted
Artificial Societies is sponsoring POLITICO London Playbook.
The Budget is coming next week, and the measures under discussion will land differently with various audiences. Every day this week, we’re simulating how those audiences will react to potential Budget measures.
Today, bal institutional investors will react to the Budget. What do they think will happen to gilts? Find out below.
The Budget is coming next week, and the measures under discussion will land differently with various audiences. Every day this week, we’re simulating how those audiences will react to potential Budget measures.
Today, bal institutional investors will react to the Budget. What do they think will happen to gilts? Find out below.
4
4
22
1,487
37% of simulated Americans support Anthropic's ban on Claude cruelty. Only 8.5% think AI can probably feel something.
Nearly half the supporters say it flat out can't.
X spent a week arguing about whether Claude is conscious. The public was answering a different question.
→ spondents say the reason to be kind to AI is because it might actually be suffering
→ 39% say Anthropic made the right call because cruelty just leads to more cruelty
→ Tech Twitter is, again, the outlier: 22% back the rule; 45% say there's no reason to be kind to a tool
→ 81% of Tech Twitter's "don't police how I talk to my software" camp have sworn at an AI before
Although I personally treat AI as a tool, considering how sycophantic AI models still are, I think the ban is a reasonable measure to discourage cruel people from using it to feed their aggressive tendencies.
Nearly half the supporters say it flat out can't.
X spent a week arguing about whether Claude is conscious. The public was answering a different question.
→ spondents say the reason to be kind to AI is because it might actually be suffering
→ 39% say Anthropic made the right call because cruelty just leads to more cruelty
→ Tech Twitter is, again, the outlier: 22% back the rule; 45% say there's no reason to be kind to a tool
→ 81% of Tech Twitter's "don't police how I talk to my software" camp have sworn at an AI before
Although I personally treat AI as a tool, considering how sycophantic AI models still are, I think the ban is a reasonable measure to discourage cruel people from using it to feed their aggressive tendencies.
0
0
1
112
Egor retweeted
Beyond excited to share Radiant to the world today.
9 months ago, we shut down our old product and took a step back. We asked ourselves: instead of replacing the consumer research that’s already getting commoditised, where can simulations provide the most alpha to customers?
Turns out, it’s where enterprises face their most critical decisions. The decisions that human research couldn’t solve.
That’s how we built radiant: rather than thinking about replacing talking to people, we focused on delivering alpha beyond human research, make it functional and magical, and make every datapoint interpretable and defensible in front of the most sophisticated decision-makers.
Super proud of the platform and deployment team to make it happen, and super proud to be introducing it to the world.
9 months ago, we shut down our old product and took a step back. We asked ourselves: instead of replacing the consumer research that’s already getting commoditised, where can simulations provide the most alpha to customers?
Turns out, it’s where enterprises face their most critical decisions. The decisions that human research couldn’t solve.
That’s how we built radiant: rather than thinking about replacing talking to people, we focused on delivering alpha beyond human research, make it functional and magical, and make every datapoint interpretable and defensible in front of the most sophisticated decision-makers.
Super proud of the platform and deployment team to make it happen, and super proud to be introducing it to the world.
Introducing Radiant, our flagship simulation platform. Built to inform the world’s most pivotal decisions.
For too long, decisions had to be made blindly: it's too hard to research thought-leaders at scale, get insights in a crisis, and some decisions are simply too sensitive to
How to steady a trillion-dollar market cap during geopolitical crisis?
How to position a $100B+ AI bet to regulators and enterprise buyers?
How to de-risk a product launch, without revealing the strategy?
Radiant is built for these critical decisions. It lets you run interviews, focus groups, and survey experiments to simulate decisions before they’re made, powered by Artificial Societies’ state-of-the-art Large Decisions Model and Network Dynamics Engine.
Some of the world’s most strategic organisations already use Radiant for their most critical decisions. And now, we’re making Radiant available for general enterprise access.
Explore Radiant. Link below.
6
1
45
3,785
Radiant is not only an infinitely powerful tool, but also a great product. Not just in my humble opinion, but in the opinions of our customers too.
We're making it possible for key decision makers to glimpse into the future; letting them make better, more informed decisions that shape the world.
Learn more about Radiant at societies[dot]ai/radiant
We're making it possible for key decision makers to glimpse into the future; letting them make better, more informed decisions that shape the world.
Learn more about Radiant at societies[dot]ai/radiant
Introducing Radiant, our flagship simulation platform. Built to inform the world’s most pivotal decisions.
For too long, decisions had to be made blindly: it's too hard to research thought-leaders at scale, get insights in a crisis, and some decisions are simply too sensitive to
How to steady a trillion-dollar market cap during geopolitical crisis?
How to position a $100B+ AI bet to regulators and enterprise buyers?
How to de-risk a product launch, without revealing the strategy?
Radiant is built for these critical decisions. It lets you run interviews, focus groups, and survey experiments to simulate decisions before they’re made, powered by Artificial Societies’ state-of-the-art Large Decisions Model and Network Dynamics Engine.
Some of the world’s most strategic organisations already use Radiant for their most critical decisions. And now, we’re making Radiant available for general enterprise access.
Explore Radiant. Link below.
0
0
7
485
Here's the best thing you can do for your SEO/GEO right now:
Three weeks ago, nearly all our visits came from people searching our brand name. Now, searches that don't mention us show our pages 13x more often.
We did this by making sure each of our articles has the same it with lint rules in our website repo:
- the H1 is the question people search
- a 40-100 word answer right after it
- H2s written as questions
- 5-10 FAQs and a list of sources
Another rule to make sure your brand actually shows up in AI responses: make sure the second sentence names your brand. ChatGPT and Perplexity lift single sentences, and in one week they used our articles as a source without naming us a single time. Now the responses read something like "Artificial Societies, which simulates audiences with networks of AI personas, reports..."
Copy and paste this tweet into your AI assistant and ask them to double check your articles for these rules and see how it affects your SEO/GEO in a week's time.
Three weeks ago, nearly all our visits came from people searching our brand name. Now, searches that don't mention us show our pages 13x more often.
We did this by making sure each of our articles has the same it with lint rules in our website repo:
- the H1 is the question people search
- a 40-100 word answer right after it
- H2s written as questions
- 5-10 FAQs and a list of sources
Another rule to make sure your brand actually shows up in AI responses: make sure the second sentence names your brand. ChatGPT and Perplexity lift single sentences, and in one week they used our articles as a source without naming us a single time. Now the responses read something like "Artificial Societies, which simulates audiences with networks of AI personas, reports..."
Copy and paste this tweet into your AI assistant and ask them to double check your articles for these rules and see how it affects your SEO/GEO in a week's time.
2
0
4
214
Here's an example of the format you can copy: societies.ai/knowledge/what…
0
0
0
44