Out of One, Many: Using Language Models to Simulate Human Samples · HackerPapers discussions

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by visarga

Thu, 08 Aug 2024 05:07:23 GMT

Reminds me of: > Out of One, Many: Using Language Models to Simulate Human Samples > We propose and explore the possibility that language models can be studied as effective proxies for specific human sub populations in social science research. Practical and research applications of artificial intelligence tools have sometimes been limited by problematic biases (such as racism or sexism), which are often treated as uniform properties of the models. We show that the "algorithmic bias" within one such tool -- the GPT 3 language model -- is instead both fine grained and demographically…

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — HN submission

Fri, 29 Mar 2024 16:50:15 GMT

Out of One, Many: Using Language Models to Simulate Human Samples (2022)

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by famouswaffles

Thu, 10 Aug 2023 18:44:44 GMT

>The difference is that an LLM isn't 1000 different intelligences. It's one intelligence, being asked to pretend to be 1,000 different people. No it's not. It does no good for a Language Model to configure a global persona. It needs to be able to predict text from wildly varying backgrounds and contexts. It's not pretending anymore than anything else it does is pretending. That's why experiments like the below actually work Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus? (https://arxiv.org/abs/2301.07543) Out of One, Many: Using Language Models to…

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by famouswaffles

Tue, 25 Jul 2023 20:31:15 GMT

Is it ? What little research we have of this kind of phenomena points towards this being very valid. Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus? (https://arxiv.org/abs/2301.07543) Out of One, Many: Using Language Models to Simulate Human Samples (https://arxiv.org/abs/2209.06899)

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by famouswaffles

Tue, 25 Jul 2023 20:06:15 GMT

Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus? (https://arxiv.org/abs/2301.07543) Out of One, Many: Using Language Models to Simulate Human Samples (https://arxiv.org/abs/2209.06899) There's been some research in this vain. To answer your question, seemingly very valid.

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — HN submission

Tue, 25 Jul 2023 17:07:12 GMT

Launch HN: Roundtable (YC S23) – Using AI to Simulate Surveys

Hi HN, we’re Mayank and Matt of Roundtable (https://roundtable.ai/). We use LLMs to produce cheap, yet surprisingly useful, simulations of surveys. Specifically, we train LLMs on standard, curated survey datasets. This approach allows us to essentially build general-purpose models of human behavior and opinion. We combine this with a nice UI that lets users easily visualize and interpret the results. Surveys are incredibly important for user and market research, but are expensive and take months to design, run, and analyze. By simulating responses, our users can get results in seconds and…

121 points; 91 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — HN submission

Fri, 21 Jul 2023 14:25:08 GMT

Show HN: Roundtable – Estimating survey results in seconds

Recent academic work ([1], [2]) has suggested that LLMs can effectively simulate different Internet subpopulations. For example, you may ask ChatGPT to emulate being a high school teacher explaining Newton’s laws of physics. Building upon this, we created Roundtable, a platform that uses LLMs to predict how people will respond to any arbitrary survey question. To do so, we needed to first reduce bias arising from GPT’s training procedure. Because these models are primarily trained on Internet data, they can be heavily skewed towards the demographics of heavy Internet users (e.g., high-income…

7 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by visarga

Sun, 15 Jan 2023 00:23:18 GMT

Related paper: Out of One, Many: Using Language Models to Simulate Human Samples > We show that the "algorithmic bias" within one such tool -- the GPT 3 language model -- is instead both fine grained and demographically correlated, meaning that proper conditioning will cause it to accurately emulate response distributions from a wide variety of human subgroups. We term this property "algorithmic fidelity" and explore its extent in GPT-3. We create "silicon samples" by conditioning the model on thousands of socio demographic backstories from real human participants in multiple large surveys…

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by visarga

Sat, 14 Jan 2023 08:30:29 GMT

Funny thing - the forum works like a language model. It doesn't have one set personality, but it can generate from a distribution of people. The language model can generate from a distribution of prompts, which might be persona descriptions. > Out of One, Many: Using Language Models to Simulate Human Samples https://arxiv.org/abs/2209.06899

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by visarga

Tue, 25 Oct 2022 22:07:53 GMT

GPT-3 has learned many identities, it's not lacking a perspective but quite the opposite, it has too many perspectives and you never know who you're talking to. Unless you prompt it with an identity. Recently it was used to run "in silico sampling" [1], a virtual poll where the respondents were GPT-3 prompted with personality backstories. So it should not be seen as an agent, it's not like us in that regard. It is the synthesis of all human culture and can assume any hat. [1] https://arxiv.org/abs/2209.06899

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — mentioned by visarga

Mon, 17 Oct 2022 09:14:05 GMT

> How? This is absolutely impossible. It is possible to use GPT-3 to impersonate someone. You give it a profile and it assumes the personality, then speaks from that point of view replicating its biases with uncanny ability. So what they did was to create a simulated poll using GPT-3 by aligning the model to personality profiles distributed identically to the real population. The result? GPT-3 can predict how an idea would poll, how a population would react to a situation. So you can simulate various policies without actually having to test in reality. You can do MCMC like AlphaGO and see ten…

Paper and discussion history. A mention is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — HN submission

Thu, 13 Oct 2022 07:57:45 GMT

Out of One, Many: Using Language Models to Simulate Human Samples

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Out of One, Many: Using Language Models to Simulate Human Samples — HN submission

Wed, 12 Oct 2022 06:17:36 GMT

Using Language Models to Simulate Human Samples

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.