Anthropic has created an AI-powered chatbot that allows users to vote for their values

In what could be the first study of its kind, AI company Anthropic has developed a large linguistic model (LLM) that has been fine-tuned for value judgments by its user community.

What does it mean for AI development to become more democratic? To find out, we partnered with @collect_intel to use @usepolis To regulate the constitution of artificial intelligence based on the opinions of about 1,000 Americans. We then trained a model against it using constitutional AI. pic.twitter.com/ZKaXw5K9sU

– Anthropic (@AnthropicAI) October 17, 2023

What does it mean for AI development to become more democratic? To find out, we partnered with @collect_intel to use @usepolis to create an AI Constitution based on the opinions of nearly 1,000 Americans. Next, we train a model using constitutional AI.

Many public-facing LLM programs have been developed with guardrails—coded instructions that dictate specific behavior—in an attempt to limit unwanted output.. For example, Anthropic’s Claude and OpenAI’s ChatGPT offer users a predefined security response to output requests related to violent or controversial topics.

However, as countless experts have pointed out, guardrails and other intrusive techniques can strip users of their autonomy. What is considered acceptable is not always useful, and what is considered useful is not always acceptable. Definitions of morality or value-based judgments can vary depending on cultures, peoples and times.

A potential remedy for this is to allow users to dictate the alignment of values ​​to AI models. Anthropic’s “Collective Constitutional AI” experiment is a stab at this “messy challenge.”

Anthropic, in collaboration with Polis and the Collective Intelligence Project, targeted 1,000 users from various demographic groups and asked them to answer a series of questions via surveys.

The challenge is to allow users to decide what is appropriate without exposing them to inappropriate results. This involves soliciting values ​​from users and then implementing those ideas into an already trained model.

Anthropics uses a method called “constitutional AI.” Driving Their efforts to fine-tune LLMs so that they are safe and useful. Guaranteed, It is about giving the model a list of rules that he must respect and training him to apply them throughout the entire process, similar to the way the constitution serves as a basic document of governance in many countries.

In the Collective Constitutional AI experiment, Anthropic attempted to incorporate collective feedback into a model constitution. Results, according to The humanistic blog post appears to have been a scientific success as much as it revealed new challenges to achieving the goal of allowing users of the LLM product to define their collective values.

One of the difficulties the team had to overcome was coming up with a new method for the measurement process.. Since this experiment appears to be the first of its kind and is based on Anthropic’s constitutional AI methodology, there is no established test to compare the baseline models with those fine-tuned using crowdsourced values.

In the end, the model that implemented data from user surveys appeared to outperform the baseline model “slightly” in terms of biased results.

According to the blog post:

“More than the resulting model, we are excited about the process. We believe this may be one of the first cases in which members of the public, as a group, have deliberately directed the behavior of a large linguistic model. We hope that communities around the world will be able to rely on the technologies Such to train culturally and contextually specific models that respond to their needs.

Explanation: The information and/or opinions expressed in this article do not necessarily represent the views or editorial line of Cointelegraph. The information provided here should not be considered as financial advice or investment recommendation. All investments and business transactions involve risks and it is every person’s responsibility to conduct due research before making an investment decision.

Investments in crypto assets are not regulated. It may not be suitable for retail investors and the entire amount invested may be lost. The services or products provided are not directed at or available to investors in Spain.