Who are you automating for?

One of the reasons I decided to publish Artificial Insights was always having an excuse to experiment with automation myself. Like most people spending significant time on their computers, over the years I have toyed with scripts, macros and code to help facilitate workflows or simply for the sake of learning.

Some readers ask me if the newsletter is written by an AI – and while it still isn’t – it makes sense to create a personal workflow optimizing for what I want to spend more (or less) time on. Writing is a form of learning, so I’m not interested in ceding this regular opportunity for synthesis to automation – at least not entirely. Having explored a few of approaches for speeding up my workflow, I’m now using Zapier to turn YouTube links using Raindrop into summarized descriptions with links in a Google Doc, and another for arbitrary links to articles. This enables tagging a link on any platform for review and editing, which helps me spend more watching and learning and less time juggling links.

This issue covers the usual array of memes, perspectives and explainers around AI with an eye towards AGI. If this resonates or if you have profiles or links to share, please always reach out.

MZ

AI is set to drastically reduce cognition costs, akin to how the internet diminished information transmission costs, according to Harvard Business School professor Karim Lakhani, an authority on workplace technology and AI. Mastering AI and machine learning is not just essential for tech-savvy employees but is crucial for all staff in a company.

Many fear that AI poses an existential risk but in this video Cleo Abram argues that AI does not have desires or intentions but may optimize for specific goals at the expense of other considerations and why it is crucial to carefully control AI systems to prevent catastrophic outcomes.

Google RT-2 is a revolutionary VLA (Vision-Language-Action) model capable of learning from both traditional databases as well as firsthand robotics data. The standout aspect is its ability to interpret new instructions, perform logical reasoning, and even respond to complex, multi-stage semantic reasoning.

**Or: HOW TO SEE LIKE A MACHINE

MoMA delves into how artists employ AI as a tool to probe existential questions around free will, machine will, and perceptions of reality. Highlighted is the "Unsupervised" exhibition for which a machine learning model churns out a blend of past, present, and future imaginings.

Threats of bias, misinformation, and irresponsible deployment of AI loom large, indicating the need for principles prioritising safety and ethics. Our way forward might be the creation of new institutions capable of responsible stewardship for AI development.

AI expert Simon Willison deep-dives into the concept and functionality of large language models and explores their challenges, including inaccuracies that may arise from relying on intuition.

We are hosting an interactive demo with Sandbox, our AI-driven co-creation tool on Tuesday 5 September. As fans of “centaur” thinking, we are developing an interactive workshop interface where participants engage with GPT prompts in a structured manner. Learn more and join our upcoming public sessions: