[HN Gopher] Ask HN: Which sources cover AI developments without ...
___________________________________________________________________
Ask HN: Which sources cover AI developments without falling for the
hype?
Popular press seems to do a bad job covering AI related
developments. Twitter is too scattered and academic papers are too
narrow. So where do working professionals get seasoned and mature
coverage of this space? What would be the AI equivalent of the
Economist, AnandTech or Tom's Hardware?
Author : pramodbiligiri
Score : 84 points
Date : 2023-11-12 17:59 UTC (5 hours ago)
| cratermoon wrote:
| At this point it's better to look for specific people rather than
| a single source. Some suggestions
|
| Gary Marcus https://garymarcus.substack.com/
|
| Temnit Gebru and Dr. Emily Bender https://www.dair-institute.org/
|
| Alex Hanna, Mystery AI Hype Theater:
| https://www.buzzsprout.com/2126417
|
| Dr. Emile P. Torres: https://www.xriskology.com/
| latenightcoding wrote:
| Gary Marcus covers Gary Marcus
| LewisVerstappen wrote:
| lol all the sources you linked have consistently been wrong at
| prredictions over the last 2 years
| croes wrote:
| If they are constantly wrong they are a good source. Just
| take the opposite
| bigfudge wrote:
| The opposite of a wrong prediction is not necessarily true.
| It depends on the form of the statement.
| jatins wrote:
| What makes Gary Marcus a "Leading expert on AI" (in his own
| words)?
| caesil wrote:
| These sources all fall for a very different kind of AI hype.
| comeonbro wrote:
| Hilarious answer
| tomalaci wrote:
| Missing Elon Musk, guy is a leading authority on all things AI
| that even the British government asks what to do about it from
| him (big /s).
| throwaway4good wrote:
| And the podcast guy that speaks really slowly ...
| jsemrau wrote:
| Substack ?
|
| AI Supremacy: https://bit.ly/3Qz8uNV
|
| Latent Space: https://bit.ly/469AAFd
|
| Encyclopedia Autonomica: https://bit.ly/3FXlVlU
|
| Deep Learning Focus: https://bit.ly/40Bi5bF
|
| Artificial Fintelligence: https://bit.ly/3SxQNAZ
| lagrange77 wrote:
| Waiting for the obligatory saint, who answers with the actual
| URLs.
| jiayo wrote:
| AI Supremacy: https://aisupremacy.substack.com/
|
| Latent Space: https://www.latent.space/
|
| Encyclopedia Autonomica: https://jdsemrau.substack.com/
|
| Deep Learning Focus: https://cameronrwolfe.substack.com/
|
| Artifical Fintelligence: https://finbarrtimbers.substack.com/
| aquir wrote:
| Unfortunately, Latent Space is not RSS friendly
| Metricon wrote:
| The RSS feed for it here appears to work:
| https://www.latent.space/feed
| swyx wrote:
| latent space author here - its a substack, it should work
|
| let me know if any other issues!
| throwaway4good wrote:
| Whats the point of the bit.ly links?
| qntmfred wrote:
| i'm guessing https://support.bitly.com/hc/en-
| us/categories/360001889231-A...
| neeleshs wrote:
| https://aisupremacy.substack.com/ https://www.latent.space/
| https://jdsemrau.substack.com/
| https://cameronrwolfe.substack.com/
| https://finbarrtimbers.substack.com/
| aquir wrote:
| AI Supremacy: https://aisupremacy.substack.com/ Latent Space:
| https://www.latent.space/ Encyclopedia Autonomica:
| https://jdsemrau.substack.com/ Deep Learning Focus:
| https://cameronrwolfe.substack.com/ Artificial Fintelligence:
| https://finbarrtimbers.substack.com/
|
| and I did not click on the links, but used a search engine to
| find the sources.
| dwaltrip wrote:
| The latent space podcast is really good. They have some top
| notch guests and ask insightful questions. They also feel
| pretty grounded compared to a lot of the hype.
| ckue wrote:
| Others already posted the resolved links, but out of interest I
| tried resolving them using ChatGPT which worked surprisingly
| well:
|
| https://chat.openai.com/share/29c10bb1-9576-43aa-9a19-e672f3...
| StableAlkyne wrote:
| Following AI-related RSS feeds from the ML research arms of the
| big companies is nice. Not as hypey as the popular press, and you
| get more technical details. Most mail clients support RSS
| nowadays, so you can even get notified when a new post is made,
| instead of periodically checking on Mondays or something. IBM,
| Microsoft, Meta (I don't think they have RSS though), nVidia,
| OpenAI, Google, and IBM all have great blogs that cover their
| work.
|
| Another good resource is the YouTube channel "2-minute papers."
| It sometimes has a lot of hype, but it does a good job of
| showcasing recent work.
| brudgers wrote:
| The term "artificial intelligence" is intrinsically hype.
|
| That's why it is always a moving bar.
|
| Good luck.
| GaggiX wrote:
| I think that when people accept that a machine can be
| intelligent, the goalpost will stop moving, but of course it
| will take time.
| nyokodo wrote:
| > when people accept that a machine can be intelligent, the
| goalpost will stop moving
|
| The goalposts will stop moving when the terms used are
| precise. What does intelligence mean? If people agree on a
| precise definition then we'll know when we've reached that
| artificially.
| GaggiX wrote:
| That would be difficult because intelligence is not black
| and white, so an arbitrary line should be decided, but at
| least everyone agrees that humans are intelligent. Maybe
| people will finally stop moving the goalpost when AI would
| be able to beat humans on every tasks.
| brudgers wrote:
| Arbitrarily calling a tail a leg doesn't give all horses
| five legs.
|
| And I wouldn't bet on AI's picking lettuce. YMMV.
| brudgers wrote:
| Intelligent machines are always in the future.
|
| First it was winning at checkers. Then it was winning at
| chess. Only recently winning go.
|
| Even your comment start with a "when" and ends with "will
| take time."
|
| The revealed wizard behind the curtain can't be unseen, the
| magic disappears and it is all just another sausage machine.
| hubraumhugo wrote:
| I had the same problem, so I ended up classifying all HN posts
| because I believe HN is my most relevant and trustworthy source
| for tech news. Example:
| https://www.kadoa.com/hacksnack/6194542d-2157-4e3c-8321-a437...
|
| This was more of an experiment for a personalizable HN feed, but
| I'll fully productize it if there is enough interest.
| Michelangelo11 wrote:
| Hey, this is really interesting. Out of curiosity -- I'm
| assuming you scrape the front page regularly, classify the
| posts, then sort them accordingly and do the summary and
| comment report via LLMs, right?
|
| In any case, I'm definitely interested in this and I can see
| myself using it fairly often.
| danielbln wrote:
| This is neat, the comment summary metrics are a nice touch. I'd
| use it.
| wheresmyshadow wrote:
| Oh man, I wish this thing had an RSS feed... would be perfect.
| hubraumhugo wrote:
| Will add RSS, good idea!
| carlossouza wrote:
| https://trendingpapers.com/
|
| Pure numbers: the top trending papers surface. They are a
| function of PageRank (citations and the importance of which
| papers cite each other), authors' previous body of work, etc...
|
| The filters help select a sub-area (NLP, Computer Vision, etc.)
| and slice what's really new (released over the last week, last 3
| months, last 6, etc.).
|
| The tool is designed to solve this problem.
| svantana wrote:
| That's a good one, and in a similar vein there's
| https://paperswithcode.com/ from facebook/meta that tracks
| papers and their accompanying github repos, along with their
| benchmark results.
| lagrange77 wrote:
| https://thegradient.pub/
|
| is not bad.
| modeless wrote:
| Twitter is good but you _have_ to filter it. Start by following
| AI researchers doing real work, not pundits. However that 's not
| enough by itself. It is essential to use the "mute words" feature
| to aggressively prune any and all tweets about politics and other
| crap. You're left with a much denser stream of people's informed
| thoughts about the latest research as it comes out.
| htrp wrote:
| How is this not a default in the featuerset?
| simonw wrote:
| I've been trying to do this on my blog. Here's my AI tag:
| https://simonwillison.net/tags/ai/
| lagrange77 wrote:
| I really like your blog!
| Michelangelo11 wrote:
| FWIW, and to answer the OP's question, I think yours and Ethan
| Mollick's blogs and social media accounts are the best sources
| for AI news right now. I've been trying to find more sources
| myself and, other than a few good but highly technical ones
| that don't focus on actual LLM use (e.g. Jeremy Howard), have
| bern coming up short.
| brianjking wrote:
| You, Ethan, & swyx are fantastic.
| jprete wrote:
| I find Zvi Mowshowitz to be the best source available
| (https://thezvi.substack.com/). He's in the X-risk camp (and so
| am I) but seems to have a clear view of some AI things being
| exciting, others being dangerous, and still others being
| irrelevant.
|
| I don't trust most AI-positive sources because they almost never
| have anything negative to say at all, so they're clearly in it to
| hype AI and not to inform anyone of true things. I don't trust
| Gary Marcus's opinion for a similar reason.
| ShamelessC wrote:
| X-risk camp?
| munchler wrote:
| Existential risk, I think.
|
| https://www.lesswrong.com/tag/existential-risk
| ShamelessC wrote:
| Ah so like a Yudkowsky type then? I don't want to lump them
| all into the same category, but since "X-risk" is given as
| a category I would just say that they're likely all very
| well educated, and for some reason deeply deeply misguided
| by some sort of philosophical hunch that is not rooted in
| actual current research but in thought experiments.
| Yudkowsky tends not to respond well or at all to criticism
| either. For real just go see the criticisms written on that
| site about his views on runaway AGI and how we get "one
| critical chance" to stop it from killing us all. His
| response is effectively "this post is very long, could you
| summarize the best bits?" and then doesn't respond any
| further.
|
| Less on topic, I'm becoming convinced that lesswrong is
| where researchers who couldn't get published go to feel
| like they got published and peer reviewed, when in reality
| they are getting peer reviewed by a massive echo chamber.
| Even less on topic and potentially downvote-inciting; it
| looks less like an echo chamber and more like a sort of
| social-media-first cult every day.
| jprete wrote:
| By X-risk person I just mean someone who thinks AI does
| have near-existential risks, and those risks are
| significant enough to merit being a top consideration in
| making decisions about AI.
|
| I don't personally base my stances on Yudkowsky's ideas,
| and never got into Lesswrong.
| trashface wrote:
| Zvi is definitely not like Yud, he's much more balanced.
| He is concerned about X-Risk but he seems to give
| reasonable space to the opposing arguments. He has a
| specific section "Other People Are Not As Worried About
| AI Killing Everyone" for this in his roundups.
|
| His roundups also cover a lot of things going on in AI
| and don't just hyperfocus on safety as the one and only
| important thing.
| mistrial9 wrote:
| humans using AI in competing teams will cause the destruction
| of ${CIV}
| capableweb wrote:
| I was looking for something similar as well since the initial SD
| launch. But I've ended up just using the HN frontpage as that,
| with good results. Most things end up here, even if not driven by
| hype, and the discussions that follow usually are "Good Enough"
| but with quite a bit of off-topic comments.
| pcf wrote:
| I follow these on YouTube:
|
| * Matt Wolfe: https://www.youtube.com/@mreflow
|
| * MattVidPro AI: https://www.youtube.com/@MattVidPro
|
| * Two Minute Papers: https://www.youtube.com/@TwoMinutePapers
|
| * Dr Alan D. Thompson: https://www.youtube.com/@DrAlanDThompson
|
| * Curious Refuge: https://www.youtube.com/@curiousrefuge
| ShamelessC wrote:
| I can't speak for the others but Two Minute Papers' general
| format oozes hype. I realize that he tends to recap on exciting
| topics, but he also sometimes "falls for" research that won't
| actually pan out for practical reasons. He also tends to cover
| things in an intensely over-optimistic way ("WOW! WHAT A TIME
| TO BE ALIVE" on literally every video...) that would have you
| think every single subject is an absolute major breakthrough
| when the reality is that many of them simply produced cool-
| looking demos that look good in a youtube thumbnail.
|
| Sorry if that's too harsh. The channel isn't all bad. The title
| of the channel is up front about being a quick synopsis that
| you can watch from time to time to keep up with the latest
| updates. That's still useful for many people. But it probably
| won't work as a good _filter_ for what will and will not be
| quickly forgotten.
|
| In fairness (again), that's a _very_ difficult challenge to
| solve. It is however the premise of the Ask HN.
| kubiton wrote:
| I watch him for a long time and he is just right.
|
| The progress on so many fields is fast.
|
| And a ton is already available to use.
|
| All the nerfs, most of Nvidia topics, etc.
|
| Or all the character animation stuff, those things are in
| games and in the industry.
|
| And even the few which show the direction make it very
| obvious were the road is going in the next 5-10 years.
|
| I'm honestly surprised that you think he is naive/to hyped.
|
| Alone the Nvidia ml denoiser and real time ray tracing is
| basically redefining graphs. We are in the middle of all of
| it.
| pixl97 wrote:
| Check out this channel if you've not seen it before
|
| https://www.youtube.com/@aiexplained-official
|
| Also robert miles, though most of his videos are older at this
| time
|
| https://www.youtube.com/@RobertMilesAI
| ShamelessC wrote:
| In general, any new research from OpenAI that is subsequently
| quickly replicated by the actual open source community (e.g.
| Latent Consistency Models for a recent example). They have put
| out research that hasn't gone anywhere in the past - which is why
| I suggest waiting until the open source community or other AI
| startups start to copy them on that specific research.
|
| Another group that is important to watch is any of the members
| from the `CompVis` group that originally developed VQGAN and
| Latent Diffusion models. Although I'm uncertain how much of the
| team remains as many seem to have realized they can do more
| research (and make some more money) by working at the various
| research labs popping up.
| jfisher4024 wrote:
| Surprised no one has mentioned AK's daily paper digest:
| https://huggingface.co/papers
|
| He also posts summaries on Twitter, or at least he used to but my
| Twitter account is glitched and I can't see Tweets anymore
| Jimmc414 wrote:
| Andrej Karpathy
|
| Simon Willison
|
| Ethan Mollick
|
| Riley Goodside
|
| Matthew Berman (for succinct howto's on YT)
| jumploops wrote:
| The recent progress in "AI" is almost entirely due to
| advancements in large language models (LLMs).
|
| In some aspects the hype is real; LLMs are extraordinarily
| performant for a wide range of previously hard tasks.
|
| On the other hand, people seem to equate these advancements with
| "strong AI" (or AGI). We are one step closer, sure, but the
| calculator was also a step forward.
|
| We've created a mirror of all (most) human knowledge, queryable
| via natural language. People look into this mirror and see
| themselves, sometimes things greater than themselves.
|
| This mirror tricks us into thinking the machine will soon replace
| us. It's so accurate, why would it not?
|
| Fortunately, it's just a mirror, and we're the bear in the woods
| seeing it's reflection for the first time. Scared and ready to
| fight.
|
| If you focus on the technology (LLMs) and throw caution at anyone
| hyping "AI" generally, you can create a filter for what's real
| and what should be questioned.
| kubiton wrote:
| There are research papers left and right.
|
| Stuff like certain architectures, reading LLM for promoting
| multi modal llms.
|
| Then we have stuff like insteuctgpt, ml models for robots, lots
| and lots of research from Nvidia for virtual simulation and
| transfer to real world, digital twin is also a relevant art in
| agi.
|
| Object detection is also much better and has nothing to do with
| llms. Segment anything from FB for example.
|
| Whisper and sd are also not LLM.
|
| There are a ton of puzzle peaces slowly falling in place left
| and right.
| pixl97 wrote:
| The issue here is "It's complicated"
|
| For example, looking at the mechanical replacement of human
| strength in the 1800 and 1900s shows people that the human
| hardship costs where real. The labor wars in the US are a good
| example of this. The process of mechanization shifted power to
| the hands of the capitalists, and was only wrestled back with
| blood.
|
| The real key of the future with AI will be the question of
| generalization. Multimodal AI does show a reasonable amount of
| ability on predicting real world events. For example, show a
| picture of a kid opening a bike and ask what is next in image
| form, and the AI will return a picture of the kid riding a
| bike. This ability of reasonable prediction based on sets of
| 'real world' input is not something that we've had in previous
| generations of computer systems. Again, if these systems
| generalize well, rapidly become cheaper, and enable the
| capitalist class to gain more wealth expect their use to
| explode at a near exponential rate.
|
| Very few reasonably educated people say "AI will never reach
| human ability", the only question that is really being asked is
| when, and in a lot of peoples eyes when has moved much sooner.
| mistrial9 wrote:
| definite "maybe" - a model can only return elements present
| in the training material. This is a powerful formula, but not
| "everything" .. blurring the story with a child, and
| prediction as a general quality.. is moving into "deception
| technique" either consciously or not IMHO
| rsynnott wrote:
| > We are one step closer, sure, but the calculator was also a
| step forward.
|
| Even that isn't particularly clear, I don't think. A
| speculative future AGI probably won't be a fancy LLM, or at
| least there's no particular reason to think it would be.
| reverse_no wrote:
| well, given that industry people and "experts" confidently
| predicted that what has happened in 2023 could happen only
| hundreds of years from now or not at all... i would say that the
| "hype" has as much veracity as anything else. the simple fact os
| that no matter how carefully you curate your news, you will have
| no idea whats coming. and coming soon. most people havent wrapped
| their head around it yet but the plain and obvious truth is that
| technology, especially AI, must be slowed, paused, regulated in
| some way because too much change too fast is very dangerous.
|
| please dont comment about previous eras of technology and change.
| nothings even comes close to comparing
| theptip wrote:
| Zvi does a very detailed roundup of the weekly news, with an AI
| x-risk lens:
|
| https://thezvi.substack.com/
|
| Pretty thorough (though verbose) if you just want to stay on top
| of developments.
| thomasweiser wrote:
| https://simonwillison.net/
| convexstrictly wrote:
| High level AI developments:
|
| https://importai.substack.com/
|
| https://www.semianalysis.com/ Hardware focused. Paywalled, but
| long article teasers contain plenty of information.
|
| Research Focused:
|
| https://codingwithintelligence.com/
|
| https://dblalock.substack.com/
|
| https://twitter.com/arankomatsuzaki
| roddylindsay wrote:
| On the AI policy side: https://aipoliticalpulse.substack.com/
___________________________________________________________________
(page generated 2023-11-12 23:02 UTC)