Boston is always a homecoming for my robotics career. I never lived here, and didn’t attend any of its 50+ institutes of higher learning, but it’s the spot where the focus of my work shifted in earnest. I had covered robots and AI previously, but it was mostly in that way gadget bloggers tended to write about the stuff, in amongst minute phone updates — check out this robot video! Isn’t it cool/scary? Don’t you want to be its friend, if it doesn’t kill you first?
As I started putting this intro together this morning, it dawned on me that we just passed the ninth anniversary of what was — in hindsight — an extremely consequential work dinner for me. If I remember correctly, it was in Cambridge, near the Akai build, where the RAI Institute has since set up shop, though admittedly, my Boston geography was even worse then than it is now.
Here’s a post from March 2017, announcing the first TC Sessions Robotics event at MIT. That happened three months later, which is to say it will be turning nine in the lead up to this year’s Automate/HRF. The dinner was an “exploratory” thing, pulling together the biggest names we could get into the same room to discuss what such an event would look like. I had two key takeaways: First, everyone in the scene knows each other. Second, they all seemed to have worked at iRobot at some point, in some capacity.
Longish story longish, my friend and then colleague, Lora, who had done much of the pulling together for said dinner, moved on to another publication, and here I am in 2026 writing a weekly thing about robots and trying to remember which Samsung Galaxy number we’re on at this point. (Lora, if you’re reading this, thanks again.)
As we head to Boston for four days of Automated podcast recording, I’m happy to report that I’ll be reuniting with several of the folks from that dinner. Of course, given everything that’s occurred in the past nine years, it’s gotten a LOT harder to get them all in one room at the same time.
If you’d like to celebrate how far we’ve come in just under a decade — or just want to say "hey" to the Automated team and some startups, please RSVP for our first-ever meetup, which is happening in just under a week. See some of you soon.
In the meantime, here's some A(s)MR to sooth your soul.
AI is remarkably good at surprisingly the humans that build it. Developers will gleefully recount stories of unexpected connections and breakthroughs they’d never envisioned until they unfolded in front of them. Such subversions of expectations are a big part of why generative AI remains such a compelling area of study. However smart and forward-thinking you might be, the system is bound to catch you off guard sooner or later.
Jagdeep Singh describes one such moment, as Rhoda worked to train robot arms to perform real-world tasks in an industrial setting. “At first glance, you might think that, okay, if you have more videos of actual workers doing the job, you might perform better,” the CEO explains. It’s a simple premise: If you want a robot to learn to perform certain tasks, focus on feeding it content of those tasks being performed.
“It turns out we found the opposite,” Singh adds. “Models that worked best were the ones that had the least curated data, where there's a bunch of other random stuff in there that again might not even involve human beings.”
Rhoda’s findings led to a reframing of how data is processed. Focusing on a specific task necessarily narrows the data input. Less curated video, on the other hand, has the potential to teach the system broader, more universal lessons.
“In hindsight, it makes sense because what you want the model to learn is just a general prior of how things move. Basically, you learn the laws of physics, intuitive physics, as you might say. And once you've learned that well, then you can pick up a new task with very little data.”
Getting there, on the other hand, will require a ton of other stuff. Rhoda’s path is lined with video — massive troves. In much the same way large language models were trained on an internet’s worth of data, the startup is pre-training its models with hundreds of millions of online videos.
“The data has been around for a long time,” says Singh. “By some estimates, we've heard 80% of all the data on the internet is video. Much of this video is openly accessible. That isn't the issue. But until recently, you just didn't have the techniques to be able to ingest all this video, to make sense of it, and then to be able to learn a prior from which you could put it in a policy prediction.”
“I forget how many tens of millions of times customers had asked Alexa to marry them or for science jokes or whatever,” Alexa and Echo VP Daniel Rausch tells me. “There's a beloved technology that doesn't just live on someone's laptop or phone. People brought Alexa into their homes and included Alexa in their lives.”
We have a firm metric for the former — or at least did as of just over a decade ago. In 2017, Amazon confirmed that more than one million people had proposed marriage to its smart assistant. Whatever motivation consumers might ultimately have, opting to devote their life to a disembodied voice emanating from a broad range of consumer electronics, it was ultimately on the system’s creators to let them down gently.
If you were among those who popped the question back then, you would have been greeted with some humorous — but firm — responses. The two of you were at very different places at that point in your lives, for example. After all, Alexa’s head, body, and everything else were stuck in the cloud.
It’s probably for the best you decided to see other people. Think of all that you’ve experienced since then — all the life you’ve lived. No hard feelings, right? Alexa’s had a glow-up in the past year, as well, as part of a fundamental rethink of how a smart assistant ought to perform — and sound — in the age of ChatGPT and its ilk. The new service, Alexa+, has since been opened to Prime subscribers in North America and the U.K., with more markets waiting to follow suit, as the company works through localization demands.
Earlier versions of Alexa had the marriage answers hard-coded into them. Customers were proposing marriage, and the assistant needed to respond appropriately, so people were tasked with writing that material. Similarly, all of those science — and every other — jokes were the product of a writers’ room.
“At the beginning, we had an incredibly talented team of comedy writers helping us build a database of jokes and responses,” says Rausch. “Now you're talking about a non-deterministic system. There are over 70 models that back Alexa, that we choose for different tasks, that we train differently, that we fine-tune differently, that we count on for different parts of the infrastructure. Really, it's about influencing the outcomes of those things, having the right training data, having the right post-training, doing the right reinforcement learning, etc. to get the experience that you're looking for that's built on sort of the tenets and foundations that you want for that experience, but is incredibly flexible and dynamic. So, it’s a very different approach to building that.”
This Wednesday, I plan to be a depressed mess for roughly 22.5 hours. But for a glorious 90 minutes, you can find me beaming like a ray of sunshine, non-alcoholic beverage in my hand at an undisclosed location in the greater Boston area. That’s right, folks — Automated is holding its first-ever happy hour. Join the podcast team, the MassRobotics crew, and assorted robots for drinks and more, from 4:30-6 PM on Wednesday, April 8. RSVP at the link below, and we’ll see you soon, before the world’s crushing realities set in, once again.
It’s a pretty good time for your autonomous warship firm to seek funding, all things considered. Saronic is showing the field how it’s done with a $1.75 billion Series D that values the company at $9.25 billion. This new funding round comes just a hair over a year after it raised $600 million at a $4 billion valuation. Cofounder and CEO, Dino Mavrookas, cited a “steady erosion of its ability to build ships and manufacture critical maritime infrastructure” as the motivating factor behind the massive funding round, which will help Saronic maintain a pace of more than 20 new ships per year, potentially alongside a new shipyard in Austin.
I’m not quite sure I’m ready to say so definitively, but I think we may have outdone ourselves with this year’s Humanoid Robot Forum lineup. As announced in an earlier edition of this newsletter, the event is happening in tandem with Automate, in Chicago on June 23-24. Seeing as how we’re still a few months out, this isn’t quite the full, 100%, set-in-stone lineup, but I’m feeling good enough about the program we’ve put together that I’m ready to share with close, personal friends, all of you.
Here’s some of who you’ll get to see speaking across two glorious summer afternoons in the Windy City: Jeff Cardenas (Apptronik), Aya Durbin (Boston Dynamics), Ali Agha (FieldAI), Nicolaus Radford (Persona AI), Amit Goel (NVIDIA), Erin McColl (Toyota Research Institute), Pras Velagapudi (Agility Robotics), David Reger (Neura Robotic), Aadeel Akhtar, (Psyonic), Elizabeth Samara-Rubio (Noble Machines), Brennand Pierce (Kinisi), and James Wells (Sanctuary AI).
We can learn a lot from snails — eat your greens, pack a hat, avoid excess salt. Above all, the little land mollusks have an important lesson to impart about moving through life slowly and with purpose. For researchers at the University of Manchester, our shelled friends serve as inspiration for a new modality of drug delivery. The work outlines a method wherein soft robots target malignant tissue, before anchoring themselves in order to deliver therapeutic drugs through a more controlled means. “Drawing inspiration from the slow, controlled and highly adaptable movements of snails and slugs,” the school writes. “The research team will mimic the animals’ unique slime-based locomotion, powered by rhythmic muscular waves and adhesive mucus, to engineer mini robots capable of navigating the gastrointestinal tract with exceptional accuracy.”
For millennia, Icarus has served as the poster boy for avoiding hubris. If the poor kid had just listened to his pops, he could have avoided a watery, wax-covered grave. Today, the enduring Greek myth lends its name to a New York-based space robotics startup, ready to hitch a ride on the International Space Station. This week, the company announced a deal with Voyager Technologies that will deliver Icarus’ Joyride robotic platform to the ISS. Set for early 2027, the mission is aimed at demonstrating and validating the robot system in a free-flying space station environment. Things should go fine, just so long as no one ventures too close to the sun.
Ranjay Krishna (Ai2/UW)- Ai2 researcher/University of Washington assistant professor, Ranjay Krishna, discusses the power of training robots in simulation.
Eric Nieves (Plus One) - Plus One's cofounder and CEO joins us backstage at the A3 Business Forum to discuss labor, humanoids, and keeping people in the loop.
Gary Cohen (iRobot) - iRobot has been a cornerstone of Boston robotics for a quarter-century. The Roomba-maker's CEO discusses its uncertain future.
The Association for Advancing Automation (A3) is North America’s largest automation trade association representing more than 1,400 organizations involved in robotics, artificial intelligence, machine vision & imaging, motion control & motors, and related automation technologies.
.png?upscale=true&width=1120&upscale=true&name=Luma%20Graphics%20%20(10).png)
.png?upscale=true&width=420&upscale=true&name=EP%2031%20Ranjay%20Krishna%20(1).png)
