Why do people work on AI when they think it could harm us long-term? Let’s not underestimate the appeal of creating beautiful things for their own sake. Also: AI used to recreating a murder victim to testify in court, the impact of AI college-cheating. In the latest Robots for the Rest of Us post.

Could #AI interpreters let us talk (and listen) to animals? More importantly, would we like what we hear? Also this week: Why economists think robots cause crime. robots4therestofus.substack.com/p/talking-to...

Good idea.

NEW: As Americans increasingly get their news from online shows and streamers, our analysis found that this expanding media ecosystem is overwhelmingly right-leaning.

A cowardly publisher is a sad sight -- even when I agree with the views of those who intimidated that publisher.

More & more I find myself treating LLMs as commodities. Some Perplexity for this, a bit of Claude for that, Grok now, NotebookLM later. With so many flooding the zone I don't feel loyal to one over another. Am I unusual?

Authoritative perspective on all the recent "robots are ready" hype.

Is there any reason for Senate Dems to vote for keeping the gov't open? Shouldn't they hold out for some concessions, like restored NIH and USAID funding? Not seeing how a shutdown is worse than letting Trump's rampage continue

Amazed people who are cheering b/c they hate today's victim of the gov't ignoring its own laws. Do people really think it could never happen to them?

At the Women's March in Manhattan today, police presence was pretty unobtrusive and chill. Except for the 8 or 9 cops standing close around a parked Tesla truck. Ready to spring to its defense, I guess.

Yet another dedicated scientist and public servant, this time from US Fish and Wildlife, showing up unpaid after she was fired to volunteer to protect an endangered species. What are we even doing here. 🎁 link: 🧪

As robots appear more frequently in ordinary lives, there are going to be times when people bash the machines. And times when robots bash people. Here’s what I think such incidents tell us about human-robot relations. (Spoiler: Nothing to do with uprisings and dystopias.)

What could possibly go wrong?

There are times in life when you don't want to give 100 % at work. When you just want to be adequate and leave on time. And maybe even make a little trouble. If this is you, don't fear #AI. It can be a great accomplice in your quest for non-excellence.

Never too early to start making summer plans.

Had a great talk with @anatperry.bsky.social about when/why people want human effort, not just a product, in their relationships. Important for deciding how and when to regulate AI therapists, AI friendship, AI romance, AI "ghosts" (recreations of deceased loved ones).

An "#AI in 2025" moment: My pdf app has a #LLM "copilot" that can summarize a document and suggest questions about it. But it can't tell me how many words are in it.

A hallmark of a state that is coming apart is when regular people have to think about the government all the time.

Also, to be specific here (because it does matter) DeepSeek is *not* open source — it’s open _weights_ with the difference being that you can run it on your own machine with the code, but you don’t know what material exactly it was trained on. An important distinction, worth emphasizing.

Well, @deepseek.bsky.social is an impressive LLM. But let's not forget that compared to other #AI, it has certain unusual limitations.

Life's tools when I was a kid were understandable (from cars to typewriters to phones etc). Now I, like most people, experience #AI and other tech as, basically, magic. I think that has some not-obvious consequences.

Can the person who stepped on the butterfly back in the Cretaceous please come forward so we can shoot you? en.wikipedia.org/wiki/A_Sound...

Data-hoovering companies already know you frighteningly well, & baking #AI tools into daily life will give corporations even more intimate data. We need a New Deal for privacy before it’s too late. Data maven WesChaar has a plan, and we had a great talk about it.

I think the cruelty is the point, actually. Performative nastiness.

Dynamically update scene graphs as an agent explores and moves around the world, in order to create robots which can function in human environments like homes. This creates a carrier-relation scene graph (CRSG) as the underlying representation, which lets it find and move to moved targets.

Congratulations to Ayana Howard @robotsmarts on her new USPTO trading card!!! www.uspto.gov/learning-and...

Some forms of concern about #AI are, in effect, toxic care — that is, care that serves the carer’s prejudices/self-interest/vanity while doing no good to the (supposedly) cared-for.

I haven't yet read the supporting data or publications. But let's boost the heck out of this. 🧪

I too find feedback aspect of LLMs is really useful for learning things and sorting through my own thoughts. Way better results come from asking "how can I do this better?" than "do this for me." Starting to worry tho about the temptation to reduce human advice-seeking, with all its risks & rewards.

A friend in real estate tried an intro-level work question on AIs (ChatGPT, Copilot, Grok, Gemini, Bing, Claude) and they all flubbed it. I tried it on some more recent products (Gemini Advanced and GPT-4o) and they got it right. Here's why I don't think that story is as simple as it sounds.

People sometimes make fun of science that sounds stupid and random. Meanwhile, a study of lizard saliva turned into a peptide medication, which was turned into a diabetes medication, which was turned into a GLP1 weight loss drug, that just became the first therapy every approved for … sleep apnea

Are engineers (like Luigi #Mangione) more prone to extremist violence than, say, poets or accountants? Some research suggests yes. Until a friend mentioned this I had forgotten that I'd written about it years ago, at @nytimes.com . #Terrorism #UnitedHealthCareShooting

Exactly.

Why a disabled person might NOT want a #humanoid #robot assistant for some tasks: Because you want people's attention on you, not the tool you're using. A point that came up in my conversation with Charlie Kemp of @hellorobot.bsky.social last month. #AI #Robotics

So I asked GPT-4o which of 2 possible fictions it'd prefer to read (because I truly am trying to decide which to try to write). It said why both were good, ie, 🤷‍♀️ . Then I asked Claude Sonnet the same question and it said, Option A is way better, and why. Way more helpful. #Anthropic wins again.

Pretty well known now that you have to get to know generative #AI, play around with it, explore it, in order to make good use of it. In this post I contrast workers who have time to do that with others, whose jobs don't give them the time. #Work #Technology #genAI

Easy to talk about robot helpers for caregiving. But how is it going to work in real life? Charlie Kemp of Hello Robot has a pretty good idea, after years developing a simple, affordable & robust little robot called Stretch. Here's our conversation. #Robots #Caregiving #Disability

When you think of a "care robot" you might imagine a machine with a face to smile, a voice to say kind words and hands to fluff a pillow. But less complicated robots make people feel capable and cared for. Plus, some other notes from last week's international conference on human-robot interaction.

Really enjoyed my conversation last week with Jonathan Hurst of Agility Robotics -- about humanoid robots in general and his in particular. And now it's up! Listen and learn why human-shaped robots are multiplying so fast in 2024.

The more amazing #AI is, the more electricity & water & "embodied carbon" it demands. OTOH AI can be (already has been) used to reduce impacts. Can its benefits make up for its own consumption? Depends on what companies and governments decided in the near future. As I try to explain here.

Baking AI into everything is going to consume a huge amount of electricity. Plus: Why it's worth worrying about AI and "enshittification" and some other news in this week's Robots for the Rest of Us.

More on that study of ChatGPT among high-end cognitive workers. Unable to go into too much detail at the NY Times due to a tight word limit, I've written a bit more about the study here:

At work ChatGPT might help you brainstorm, mislead you on facts and figures & make you question the meaning of your existence, a recent study suggests. I describe it here, in today's @nytimes.com :

Getting tired of people praising boredom as a good thing (it sparks creativity!). That requires freedom. Without it, boredom is not "nothing to do." It's "something to do" where the "something" is soul-killing. As children and prisoners know.