[HN Gopher] Automating the search for artificial life with found...
___________________________________________________________________
Automating the search for artificial life with foundation models
Author : hardmaru
Score : 145 points
Date : 2024-12-24 02:47 UTC (20 hours ago)
(HTM) web link (sakana.ai)
(TXT) w3m dump (sakana.ai)
| wintercarver wrote:
| Congrats David & the whole team! Really enjoy everything Sakana
| AI produces and always look forward to your research results.
| hamburga wrote:
| I actually found artificial life. Crocs. They keep on reproducing
| effectively and walking around (symbiotically with humans), with
| some mutation though the polysexual recombination process of
| Product Manager design reviews.
| vintermann wrote:
| I think that it's a bit silly to call something life just
| because it resembles stuff you see under a microscope.
|
| But I can't deny that it's beautiful. Unlike crocs.
| mvkel wrote:
| Fun fact: Sakana AI is founded by some of the authors of the
| original transformer paper, "Attention Is All You Need"
| sourcepluck wrote:
| Where's the fun part? I can't exactly imagine throwing this out
| as an anecdote to entertain a few friends during a
| sophisticated little soiree.
| vintagedave wrote:
| No need to poke fun. I found it interesting. Among friends
| who are interested in AI it's the kind of random fact you'd
| throw into conversation.
| jazzyjackson wrote:
| Some people have a low bar for fun, for example, learning
| something new that connects to something they already knew,
| and saying to themselves, "Neat!"
| diggan wrote:
| > to entertain a few friends during a sophisticated little
| soiree
|
| Isn't this basically what (we'd like to think) HN is?
| rbanffy wrote:
| Minus food and drinks, and in-person interaction. It's sad
| that joining a Teams/Zoom/* channel feels so much like work
| these days.
|
| Maybe if we could make it feel more like house parties it'd
| work.
| fedeb95 wrote:
| try harder
| exe34 wrote:
| you have friends?
| jeroenvlek wrote:
| In this context I can also highly recommend the Sara Walker
| episodes on Lex Fridman:
|
| https://youtu.be/-tDQ74I3Ovs?si=1m0JV8gZEl4WFedG
|
| https://youtu.be/SFxIazwNP_0?si=R7yZroSNbw5Jjc0H
|
| https://youtu.be/wwhTfyX9J34?si=ceXh_aehsjQPklUT
| nextworddev wrote:
| Curious - what's the intended product direction of Sakana AI? Is
| it mainly a research lab or is it doing commercialization?
| diggan wrote:
| It's a incorporated for-profit company with VC investments, so
| somewhere/somehow there needs to be commercialization.
| rbanffy wrote:
| Can always be a pure-play IP house.
| ribadeo wrote:
| The name of this company has real meaning in Portugues which I
| reckon is unintended.
| jazzyjackson wrote:
| I mean it's Japanese for Fish, but yeah, perhaps we need a
| database of false cognates sorted by number-of-languages-that-
| consider-it-vulgar
|
| As for Portuguese, GPTo3 tells me "depending on context it can
| mean "bastard," "scumbag," "dirty-minded jerk," or imply that
| someone is a lecherous creep. It's essentially an insult
| calling someone sleazy or untrustworthy."
|
| Would you say that's about right?
| diggan wrote:
| > perhaps we need a database of false cognates sorted by
| number-of-languages-that-consider-it-vulgar
|
| Or, like most people, we can assume the intent from the
| context and if someone says "Use git", we know they're not
| telling us to use a bum/rat/scum/whatever but the SCM :)
| Y_Y wrote:
| Use bum
| theGnuMe wrote:
| Are these cellular automata or Something more?
| readyplayernull wrote:
| 2025 prediction: Wolfram declares agentic cellular automaton
| supremacy.
| fedeb95 wrote:
| he's slowly building his army that will conquer all the
| computers in the world.
| upghost wrote:
| Wow, it is _really_ interesting the difference in comments
| between ALife and AI stories on HN.
|
| For some of you out there, there's a great book that really
| hasn't gotten enough attention called "The Self-Assembling Brain"
| [1] that explores intelligence (artificial or otherwise) from the
| perspectives of AI, ALife, robotics, genetics, and neuroscience.
|
| I hadn't realized the divide was a sharp as it is until I saw the
| difference in comments. i.e. this one[2] about GPT-5 has over
| 1000 comments of emotional intensity while comments on OP story
| are significantly less "intense".
|
| The thing is, if you compare the fields, you would quickly
| realize that which we call AI has very little in common which
| intelligence. It can't even habituate to stimuli. A little more
| cross disciplinary study would help is get better AI sooner.
|
| Happy this story made it to the front page.
|
| [1]: https://a.co/d/hF2UJKF
|
| [2]: https://news.ycombinator.com/item?id=42485938
| fedeb95 wrote:
| thanks for your resources. I am myself concerned with the
| question of artificial life, and I wonder if it is even
| possible to search for it, or rather it will emerge on its own.
| Perhaps, in a sense, it is already emerging, and we humans are
| its substrate...
| upghost wrote:
| I'm not even sure that the goal of Artificial Life is
| actually "life", although that may be the AGI equivalent of
| ALife -- AGL or "Artificial General Life"?. In practice I
| think the discipline is much closer to the current LLM hype
| around "Agentic AI", but with more of a focus around the
| environment in which the agents are situated and the
| interactions between communities of agents.
|
| Much like the term "Artificial Intelligence", the term ALife
| is somewhat misleading in terms of the actual discipline.
|
| The overlap between "agentic AI" and ALife is so strong it's
| amazing to me that there is so little discussion between the
| fields. In fact it's closer to borderline disdain!
| Y_Y wrote:
| Apart from the obvious distinction that many of us on HN are
| making (or trying to make) money on LLMs I think you've also
| hit a broader point.
|
| There appears to be a class of article that have a relatively
| ratio of votes to comments, and concerns such topics as, e.g.
| Programming Language Theory or high-level physics. These are of
| broad interest and probably are widely read, but are difficult
| to make a substantial comment on. I don't think there are knee-
| jerk responses to be made on Quantum Loop Gravity, so even
| asking an intelligent question requires background and thought
| and reading the fine article. (Unless you're complaining about
| the website design.)
|
| The opposite is the sort of topic that generates bikeshedding
| and "political" discussion, along with genuine worthwhile
| contributions. AI safety, libertarian economics, and
| Californian infrastructure fall into this bucket.
|
| This is all based on vibes from decades of reading HN and its
| forerunners like /. but I would be surprised if someone hasn't
| done some statical analyses that support the broad point. In
| fact I half remember dang saying that the comments-to-votes
| ratio is used as an indicator of topics getting too noisy and
| veering away from the site's goals.
| rbanffy wrote:
| > many of us on HN are making (or trying to make) money on
| LLMs
|
| I'd also highlight the misalignment between creating better
| AI and working towards AGI and extracting value right now
| from LLMs (and money investors).
| j0hnb wrote:
| Before I read the article all I could think about was what if AI
| was used with SETI's data, would we find something there?
| rrr_oh_man wrote:
| What has prompted you to comment, after almost 10 years?
| j0hnb wrote:
| No specific reason, i'm here multiple times a day but I
| rarely comment.
___________________________________________________________________
(page generated 2024-12-24 23:01 UTC)