Ethics of AI agents is a topic that we need to begin grappling with
Prabhakar played a major role in the passage of President Biden’s executive order in 2023 on AI, which sets out rules for tech firms to make AI more transparent and safer (although it depends on voluntary participation). She held several government positions before joining President Biden’s Cabinet, including rallying for the domestic production of semiconductors and heading DARPA, Pentagon’s famed Research Department. I had the chance to meet Prabhakar this past month. The change in administration comes at a chaotic moment for AI. Trump’s team is yet to present a clear theory of how they will deal with artificial intelligence. However, many people within the team want the executive order revoked. Trump stated this in July when he endorsed the Republican Platform, which said that the executive order was “a hindrance to AI innovation” and that it imposed radical leftwing ideas on technology development. Elon Musk
will complicate that narrative. He has for years expressed concerns about doomsday AI situations and supported some regulations aimed at promoting AI safety. Prabhakar’s insights on what has happened to date are fascinating.
Deeper LearningThese AI Minecraft characters did weirdly human stuff all on their ownThe video game Minecraft is increasingly popular as a testing ground for AI models and agents. This is a trend that Altera has recently adopted. Up to 1000 software agents were released at once, each powered by large-scale language models. They developed an impressive range of personality traits, specializations, and preferences without any further input from their creators. They were able to make friends, create jobs and spread religion on their own. It is yet another example of the bizarre and impressive behavior of AI agents who are not prompted by humans. Agents are a big deal to those who want them. Robert Yang, Altera’s CEO, sees the Minecraft experiment as a first step to large-scale “AI civilisations” where agents can coexist and collaborate with us in digital space. Yang says that the true power of AI is unlocked once we have autonomous agents who can work together at scale. Niall Firth shares more.Bits & Bytes
OpenAI explores advertising
