118 karma · joined November 10, 2023
It's much faster to read the article than watch the video, even though that hurts him by 1 view.
I just watched parts of the video after reading because I wanted to see his explanations.
One of the few really good creators out there.
So the penalty for breaking the law is being forced to follow the law for 5-20 years? (Or am I simplifying it too much and missing something?)
I'm recalibrating my own behavior to upvote more.
Is it the desired behavior of HN that silent upvotes are for agreement? (Instead of a positive comment that doesn't add substantially to the discourse?)
More informative, less clickbaity.
Other than that, I quite enjoyed the article and the parallels to how modern cities might shift with changes in climate patterns.
The hypothesis that they propose is that 'the addition of an unusual character like Batman on the subway made people who otherwise may have been lost in their own thoughts or podcast look around and be more aware of people around them, and this more likely to notice the pregnant woman and offer their seat'
I'm looking forward to the follow up paper to test the hypothesis.
I see ports on its edges that look like fluid connectors.
I'm not too sure how I activated it, but on one occasion it did launch into a search of my history when I was trying in the address bar.
It looks like Google is just silently adding the useful features into their AI. Product (like they did with the ai overviews)
I found though that yielding just one response per side seemed pretty weak, especially when it's not always picking the best arguments for each side. (And not returning enough of the passage to fully show the argument)
It doesn't sound thorough but rather just like it's picking the first vaguely similar set of lines that it could find.
I wonder if subtle tweaks to your prompts could make it significantly more convincing?
I had a fair understanding of passkeys before reading your response, but I too learnt something from it.
Thank you.
The article proposes a new word 'architecture debt' for most of what people consider 'tech debt', and then tries to say that architecture debt is a more serious concern than tech debt.
Most of the examples used to justify creation vs consumption can also be explained by low scale vs high scale (cost sensitive at high scale) or portability.
I invested in two wireless handheld keyboard+pointer inputs to match the different input styles of me and my wife.
Completely bypasses all ads with less effort than setting up a pihole or torrent+Plex server, and the bonus is avoiding the surveillance from the TV's 'consumption OS'
Does it move a mouse pointer on the screen?
Sad to hear that I'm in a minority for loving that input.
We're also building out a technical steering committee to help guide our direction on topics like this. Safety is a big category where having direction from across the community will be important.
We're using websockets as the interface between the server and the robot itself, which to the best of my exploration does have the ability for simultaneous connections.
So far, I've observed that for Claude and Gemini, which are what we've been testing most with, the Language model has been pretty good at recognizing its a faulty initial interpretation when it queries more information from the system.
Running out of tokens is a more significant issue. We saw it a lot when we queried imaged topics, which led us to try writing better image interpreters within the MCP server itself (credit to my collaborators at the Hanyang university in Korea) to defend the context window. Free tiers of the language models also run out of tokens quite quickly.
PS - Thank you for the questions, I'm enjoying talking about this here on HN with people who look at it critically and challenge us!
This is the video of interacting with and debugging an industrial robot. (A few of the other comments here have been talking about this, that we see some amount of what looks like emergent behavior) https://www.youtube.com/watch?v=SrHzC5InJDA
This is a video from a collaborator research lab controlling a Unitree Go (robot dog) https://youtu.be/RW9_FgfxWzs?si=o7tIHs5eChEy9glI
The demo video with the Unitree Go (robot dog) uses this approach to give the LLM additional context of the custom poses available to it.
The underlying stack definitely can connect to multiple robots simultaneously. Our current implementation is sequential, where the language model can connect from one robot to the next (and then back).
But it is definitely possible for us to write it to be simultaneous as well.
For the industrial robot (in the video on the main readme page) I intentionally gave Claude no context beforehand. All of the inferences that you see there are based on information that it got through the MCP tools that let it analyze the ROS topics and services on the robot.
In fact, I had a starting prompt to ignore all context from previous conversations because this looked to be like an example of emergent behavior and I wanted to confirm that it was not picking things from my earlier conversations!