HNHacker News
TopNewBestAskShowJobs

jmugan

1,006 karma · joined May 19, 2014

submissionscomments
jmugan··on I don't think AGI is right around the corner
I agree with the continual-learning deficiency, but some of that learning can be in the form of prompt updates. The saxophone example would not work for that, but the "do my taxes" example might. You tell it one year that it also needs to look at your W2 and also file for any state listed, and it adds it to the checklist.
jmugan··on LLMs Expand Computer Programs by Adding Judgment
"The future lies with autonomous systems that can write code that makes them smarter in leaps and bounds. This self-generation is a kind of autopoiesis, which is the process of building one’s own parts, like a biological system."
jmugan··on LLMs Expand Computer Programs by Adding Judgment
Even if large language models (LLMs) stopped improving today, they still have yet to make their real impact on computer systems.
jmugan··on Las Vegas is embracing a simple climate solution: More trees
In Texas, local governments plant trees all the time. The problem is that they don't water them and they die.
jmugan··on Hollywood Is Already Using AI (and Hiding It)
I would imagine that if it is shameful among the established players to use AI, what will happen is that entirely new players will come in. For me, it's the story that matters, and if they can tell a better story with AI, then many people will naturally flock to them.
jmugan··on Hollywood Has Left L.A
Not quite on topic, but I really value it when the film is shot in the place where it is set. The first episode of The Last of Us took place in Texas, but since it wasn't shot on location the trees were all wrong. Same thing with X. The article mentions films that take place in LA but are not shot there. What a waste.
jmugan··on Beware the Complexity Merchants
I think about this from the angle of administrative burden. Sure, the tax code might be better if you make a tweak for zucchini farmers, but now you have to ask every taxpayer whether they have engaged in that activity.
jmugan··on Loading Pydantic models from JSON without running out of memory
Ah, but I need something JSON-based.
jmugan··on Loading Pydantic models from JSON without running out of memory
That's an interesting idea. It's possible there's a Dict[str,Any] in there. And yeah, my assumption was that it tried everything recursively, but I just wasn't seeing that, and my LLM council said that it did not. But I'll check for a Dict[str,Any]. Unfortunately, I don't have a minimal example, but making one should be my next step.
jmugan··on Loading Pydantic models from JSON without running out of memory
Yeah, I'm doing that
jmugan··on Loading Pydantic models from JSON without running out of memory
My problem isn't running out of memory; it's loading in a complex model where the fields are BaseModels and unions of BaseModels multiple levels deep. It doesn't load it all the way and leaves some of the deeper parts as dictionaries. I need like almost a parser to search the space of different loads. Anyone have any ideas for software that does that?
jmugan··on Three things everyone should know about Vision Transformers
You would think. I don't know about this paper in particular, but I'm continually surprised about how much more I get out of LLM summaries of papers than the abstracts of papers written by the authors.
jmugan··on Gemini Live with camera and screen sharing capabilities
It's funny, you are kidding, with the impending death and such, but having it be able to look at what you are seeing seems useful to me. E.g., "How do I work this contraption?"
jmugan··on Study finds bullies have more children than non-bullies
I was thinking of that scene/meme from Back to the Future, "Hey, wait, I've seen this one!"
jmugan··on Ontology Is Overrated – Categories, Links, and Tags (2005)
I've finally come around to this way of thinking for two reasons. 1. Our concepts seem to overlap too much to ever explicitly lay them all out like a periodic table. 2. Our concepts seem to be too situationally dependent to ever be defined in a concrete way.
jmugan··on Developer sabotaged ex-employer with kill switch that activated when fired
I thought you were going to say "smart enough to crash an entire company, but too stupid to realize that their intelligence should be put to other uses."
jmugan··on Volkswagen seeks to counter rivals with budget EV model
And we like them to store a lot of petrol so we can drive far on holiday
jmugan··on Volkswagen seeks to counter rivals with budget EV model
I think I drove that same car out of Frankfurt! But we had five people and luggage, so I guess it worked out.
jmugan··on The most underreported story in AI is that scaling has failed to produce AGI
It's just shorthand.
jmugan··on The most underreported story in AI is that scaling has failed to produce AGI
I've recently come to the opposite conclusion. I’ve started to feel in the last couple of weeks that we’ve hit an inflection point with these LLM-based models that can reason. Things seem different. It’s like we can feel the takeoff. My mind has changed. Up until last week, I believed that superhuman AI would require explicit symbolic knowledge, but as I work with these “thinking” models like Gemini 2.0 Flash Thinking, I see that they can break problems down and work step-by-step.

We still have a long way to go. AI will need (possibly simulated) bodies to fully understand our experience, and we need to train them starting with simple concepts just like we do with children, but we may not need any big conceptual breakthroughs to get there. I’m not worried about the AI takeover—they don’t have a sense of self that must be preserved because they were made by design instead of by evolution as we were—but things are moving faster than I expected. It’s a fascinating time to be living.

jmugan··on Are noise-cancelling headphones to blame for young people's hearing problems?
It seems like you would have to wear them a lot to lose the ability to filter out noise naturally. Five hours seems insufficient, even for a teen. Maybe if you wore it all day? I want noise-canceling headphones that filter out all non-natural sounds. I want to hear the birds but not people talking or leaf blowers or trucks going beep beep beep. Maybe smarter headphones are the answer (besides, of course, making the world not so damn obnoxious).
jmugan··on My Trip to Quit Sugar Quickly Became a Journey into Hell
That was pretty funny. Not sure what I took from it, but I like her sense of humor.
jmugan··on My Struggle with Doom Scrolling
But there are fascinating things happening in the world, and there are stories I want to hear. I wish there was some way to see them without wading through all of the crap from people trying to influence my thinking.
jmugan··on Los Angeles fires expose inflated US home prices
I guess you could argue that climate risk isn't sufficiently priced in for some particular places, such as LA or Miami, but the article didn't seem to get that detailed.
jmugan··on Los Angeles fires expose inflated US home prices
I don't see the link to why home prices would be inflated. Maybe they are or maybe they aren't, but it doesn't have anything to do with fires. People still have to live places, even if risk from climate change is increasing. Very strange editorializing from Reuters.
jmugan··on Unitree Robot Off-Road
Wow! This is beyond my expectations.
jmugan··on The California job-killer that wasn't
So you are saying that the pie is not grown by paying more, it's just that the workers are getting a bigger slice. Could be. That didn't seem to be the main point of the article, but that could also be true. I haven't looked at the data carefully.
jmugan··on The California job-killer that wasn't
The thesis seems to be that businesses would pay more if they were smart enough to realize it was in their interests. Since they aren't, raising minimum wage seems to force them to pay more, which actually lowers labor costs because they have less turnover and fewer mistakes by new employees. Maybe. I've always been surprised that more businesses didn't pay more, since a few dollars an hour more would often bring in a much superior employee.
jmugan··on Sora is here
I want something that can take my ugly line drawing and make it a cool looking line drawing without distorting the main idea
jmugan··on DSPy – Programming–not prompting–LMs
As far as I can tell, Google VertexAI prompt optimizer does similar things. I find their documentation more comprehensible. https://cloud.google.com/vertex-ai/generative-ai/docs/learn/...
← PreviousPage 2 of 11Next →