Month: April 2023

“Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day,” write the researchers in their paper… To pull this off, the researchers relied heavily on a large language model for social interaction, specifically the ChatGPT API. In addition, they created an architecture that simulates minds with memories and experiences, then let the agents loose in the world to interact…. To study the group of AI agents, the researchers set up a virtual town called “Smallville,” which includes houses, a cafe, a park, and a grocery store…. Interestingly, when the characters in the sandbox world encounter each other, they often speak to each other using natural language provided by ChatGPT. In this way, they exchange information and form memories about their daily lives.

When the researchers combined these basic ingredients together and ran the simulation, interesting things began to happen. In the paper, the researchers list three emergent behaviors resulting from the simulation. None of these were pre-programmed but rather resulted from the interactions between the agents. These included “information diffusion” (agents telling each other information and having it spread socially among the town), “relationship memory” (memory of past interactions between agents and mentioning those earlier events later), and “coordination” (planning and attending a Valentine’s Day party together with other agents)…. “Starting with only a single user-specified notion that one agent wants to throw a Valentine’s Day party,” the researchers write, “the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time….”

To get a look at Smallville, the researchers have posted an interactive demo online through a special website, but it’s a “pre-computed replay of a simulation” described in the paper and not a real-time simulation. Still, it gives a good illustration of the richness of social interactions that can emerge from an apparently simple virtual world running in a computer sandbox.
Interstingly, the researchers hired human evaluators to gauge how well the AI agents produced believable responses — and discovered they were more believable than when supplied their own responses.

Thanks to long-time Slashdot reader Baron_Yam for sharing the article.

What they tell me is obvious to anyone watching. Competition is forcing them to go too fast and cut too many corners. This technology is too important to be left to a race between Microsoft, Google, Meta and a few other firms. But no one company can slow down to a safe pace without risking irrelevancy. That’s where the government comes in — or so they hope… [A]fter talking to a lot of people working on these problems and reading through a lot of policy papers imagining solutions, there are a few categories I’d prioritize.

The first is the question — and it is a question — of interpretability. As I said above, it’s not clear that interpretability is achievable. But without it, we will be turning more and more of our society over to algorithms we do not understand… The second is security. For all the talk of an A.I. race with China, the easiest way for China — or any country for that matter, or even any hacker collective — to catch up on A.I. is to simply steal the work being done here. Any firm building A.I. systems above a certain scale should be operating with hardened cybersecurity. It’s ridiculous to block the export of advanced semiconductors to China but to simply hope that every 26-year-old engineer at OpenAI is following appropriate security measures.

The third is evaluations and audits. This is how models will be evaluated for everything from bias to the ability to scam people to the tendency to replicate themselves across the internet. Right now, the testing done to make sure large models are safe is voluntary, opaque and inconsistent. No best practices have been accepted across the industry, and not nearly enough work has been done to build testing regimes in which the public can have confidence. That needs to change — and fast.
The piece also recommends that AI-design companies “bear at least some liability for what their models.” But what legislation should we see — and what legislation will we see? “One thing regulators shouldn’t fear is imperfect rules that slow a young industry,” the piece argues.

“For once, much of that industry is desperate for someone to help slow it down.”