A very gentle introduction to large language models (without the hype)

🌻 Today is the vernal equinox, aka the first day of spring in the northern hemisphere. The days are only getting longer from here! (Until the solstice on June 20)
Today: AI fears, Figma tips and tricks, and the psychology behind impulse shopping. By Harris Sockel

Why are we, as humans, so afraid of what we don’t understand?

Take artificial intelligence: Last year, Elon Musk and 33,000 others signed an open letter calling for a six-month pause in developing systems smarter than GPT-4. The letter’s signatories included politicians (Andrew Yang), technologists (Steve Wozniak), and ordinary people. It summarized many of the ambient fears of AI I’ve been reading about on Medium and elsewhere. They all boil down to one basic fact: We don’t know what the future of AI will look like, so let’s slow down.

Of course there are very valid reasons to build commonsense stopgaps in AI development. Last week, a group of scientists agreed not to use AI to develop bioweapons (makes sense). But fear usually comes from a place of confusion, and I recently found two stories on Medium that helped me (and will hopefully help you) understand AI in new ways:

• To give you a basic sense of what LLMs are and how they work, here’s a very gentle introduction to large language models (without the hype) from Mark Riedl, computer science professor at Georgia Tech. He uses the metaphor of a giant typewriter for an LLM, where every key is a word. Because LLMs are trained on the internet, the takeaway here is: “If we were to think about what an LLM is trained to do, it is to produce text that could have reasonably appeared on the internet.”
• Vincent Vanhoucke, Director of Robotics at Google DeepMind, recaps a recent conference on generative AI. “There was a sense of change in the air,” he says. “Many presentation systems showcased robot systems that actually… kind of worked!” And many of these robots ran on LLMs. Not in the simplistic way of “they could talk to you” but in the deeper sense: They used LLMs to plan their actions. “LLMs know simple truths such as ‘a book belongs on a shelf, not in a bathtub,’ or ‘how one would go about making coffee,’ and that turns out to be a big deal for embodied agents trying to act in the real world. So it’s no surprise that the first corner of robotics to be affected by LLMs would be planning.”

What else we’re reading

• Figma, the tool in which designers create many of the apps you use daily, released a new feature two weeks ago: Multi-editing now allows designers to edit several components at the same time. Former Figma employee Joey Banks breaks down all the shortcuts.
• “There was never a time in my childhood where decisions were thought out with care, or in a rational manner. Decisions were made spontaneously; recklessly,” writes poet Kathy Parker in an essay about peeling back the layers of her childhood trauma to uncover why she is so terrible at decision-making. Imitation shapes some of our earliest behaviors, she writes, which can set us on paths that determine the course of our lives long-term.

Your daily dose of practical wisdom: about psychology

To avoid impulse shopping, create a “shopping rubric” — a list of criteria for what you will and won’t buy — and hold yourself accountable to it.

first morning of spring in nyc, 8:10am today

Written by Harris Sockel
Edited and produced by Scott Lamb & Carly Rose Gillis

Every weekday, the Medium Daily Edition offers useful, human perspectives to help you become a better reader, writer, and thinker.

Did someone forward you this email? Sign up here.

Want to browse the best of Medium? Explore staff picks.

Unlock a library of unlimited knowledge and insight with a Medium membership.

Questions, feedback, or story suggestions? Email us: tips@medium.com