> vice president of commercial sales, said, “If someone can just show us how to replicate [these uncommanded discharges] we will absolutely look at this from all aspects to make sure there isn’t any truth to this.
Am I reading it correctly: if you provide us with a detailed bug report we will make sure to properly justify why we think it is not a bug and "works as intended"? :)
Is there a fundamental reason why a "perfect mechanism" like this sounds just like a "perfect string" = midi guitar instrument = harmonic sequence of a vibrating string? Does it mean that human fingers are responsible for the remaining harmonics that we associate with the "guitar sound" on top of the "theoretical guitar tembre".
"Our key innovation is a new collection of datasets called PixMo that includes a novel highly-detailed image caption dataset collected entirely from human annotators using speech-based descriptions, and a diverse mixture of fine-tuning datasets that enable new capabilities. Notably, PixMo includes innovative 2D pointing data that enables Molmo to answer questions not just using natural language but also using non verbal cues. We believe this opens up important future directions for VLMs enabling agents to interact in virtual and physical worlds. The success of our approach relies on careful choices for the model architecture details, a well-tuned training pipeline, and most critically the quality of our newly collected datasets, all of which we have released."
This is a solid engineering project with a research component - they collected some data that ended up being quite useful when combined with pre-existing tech. But this is not rocket science and not a unique insight. And I don't want to devalue the importance of solid engineering work, but you normally don't get paid as much for non-unique engineering expertise. This by no means sounds unique to me. This seem like a good senior-staff research eng project in a big tech company these days. You don't get paid 250M for that kind of work. I know very talented people who do this kind of work in big tech, and from what I can tell, many of them appear to have much more fundamental insight and experience, and led larger teams of engineers, and their comp does not surpass 1-2M tops (taking a very generous upper bound).
A PhD dropout with an alright (passable) academic record, who worked in a 1.5-tier lab on a fairly pedestrian project (multimodal llms and agents, sure), and started a startup.. Reallyttrying to not sound bitter, good for him, I guess, but does it indicate that there's something really fucked up with how talent is being acquired?
Hm, I thought that these salaries were offered to actual "giants" like Jeff Dean or someone extremely knowledgeable in the specifics of how the "business side" of AI might look like (CEOs, etc). Can someone clarify what is so special about this specific person? He is not a "top tier athlete" - I looked at his academic profile and it does not seem impressive to me by any measure. He'd make an alright (not even particularly great) assistant professor in a second tier university - which is impressive, but is by no means unique enough to explain this compensation.
Ok, now give me an example of a resource manager (eg in a game) that has methods for loading resources into memory and also for releasing such resources - all of a sudden if a system needs to give away pointer access to its buffers, things become more complicated and arena allocators are not enough.
As another example, I was treated horribly by a ex-spouce for years, and for some reason I didn't leave, and only covered up for my spouce's bad behavior. Retrospectively it was destructive, but in the moment it felt like the right thing to do (in a very roundabout way). I don't want this to happen again, so I'm trying to understand and catch behaviors that got me there in the first place.
But if we drop the false nostalgia and think of about the overall "we think too much and feel too little" sentiment - i can relate to that.
Right, it is bizzare to read that someone "unearthed a forgotten AI technique" that you happened to have worked with/on when it was still hot - when did I become a fossil? :D
Also, if we're being nitpicky, diffusion model inference has been proven equivalent to (and is often used as) a particular NF so.. shrug
TIL about several more cool gemini-powered prototyping tools: both 1) Canvas tool option in Gemini web (!) app and 2) Build panel in Google AI Studio can generate amazing multi-file shareable web apps in seconds.
Always wanted to experiment with a syntax like this myself, thanks to author! I completely agree with you reasoning re complexity of mental model for indexing vs broadcasting. Moreover, it appears to me that such a representation should allow for finding more optimal low level impls (something like deeper "out of order" op fusing, idk)? I've seen a paper from either nvidia or meta around five years ago doing exactly that - translating an index-based meta-language built on top of python into cuda kernels (usually several variants and picking the best), can't find the reference unfortunately.
I don't know APL, but that has been my thought as well - if APL does not offer much over numpy, I'd argue that the I'd argue that later is much easier to read and reason through.
In fact, I also converge to sql as a universal data transformation language. External analogs include things like duck db. Unfortunately, even with pipe syntax sql lacks expressiveness causing me to revert to c-style macros in sql (eg making table name dynamic), which in the long run makes things far less maintainable if anything.
I work in one of large tech companies, and I can attest that while the idea seems very neat in theory (esp if your schemas are typed), and even if you define an api for defining new building blocks, sooner or later people realize that they need to dynamically adjust parts of the pipeline, and they write components to dynamically set and resolve these, and then other components on top of these components, and then complements for composing components - and now you forced yourself into implementing a weird and hard to debug functional programming language in yaml which is not a place someone wants to find themselves in :'(
one lesson I learned from this: any bit of logic that defines a computation should prefer explicit imperative code (eg python) over configuration, because you are likely to eventually implement an imperative language in that configuration language anyway
Makes sense. How can you even approximately estimate higher order differences in conditional moments in such a high dim space? Seems statistically impossible to get a reasonable estimate for a gradient. Moment matching in sample space has always been very hard.
Not all sexually explicit content is harmful in all contexts for sure, but in many contexts it is fairly universally considered harmful (eg content involving minors). Do you have means of distinguishing between the two? Are you suggesting that a company must invests millions into teaching the model where exactly the red line lines so that it can have a conversation close to it but without crossing it? Or you suggest biting the bullet and releasing the model not only capable of generating eg child porn, but also having a >0 chance of randomly discussing it in unrelated contexts? Chance of error is always there, and companies decided that a risk of really bad behavior in benign context overweights the gains. Imho, a decision to not play whack a mole with this land mine is quite rational, esp considering gains vs risks vs costs. Think of it as a cost cutting measure, not as an infringement on free speech. You are free to invest you own money into this problem if you think that's a grave mistake and a missed opportunity. The first project to push the automated generated content moderation against what is considered appropriate in the given context far enough to make it economical for companies to put their guard down could actually be worth a lot if you think there's market for it (eg agents on dating websites? idk, you tell me)
Same. You can clearly tell how much better of an author she got by the time the Psalm came out. It is a very solid book. The second one in the series seemed more confused, but the first one felt very thought though and intentional.
Oh, I don't think it is the US as a whole that's meant to benefit from this coercive behavior. And it's not just the foreign policy. While Trump keeps pointing fingers at those "who are to blame for our problems" (transes stealing our children, biden stealing elections, immigrants stealing jobs and welfare, Ukrainians stealing aid) and the society stays divided on these issues, plenty of opportunities for business (extortion) arise for himself and his buddies, especially now that his hands are not tied by anything. This strategy worked wonders for another narcissist bully thug named Putin. That's basically divide and conquer.
Not everyone benefits from cooperation. Feuds are a great "business opportunity" - while everyone's divided and at each other's throat, you can take them out (or extort/bully) one by one, without any consequences, especially if you have the biggest and scariest stick in the block. So, solidarity, cooperation, and people sticking up for each other are absolutely awful for thugs and bullies.
More death wouldn't, but a solid established and supported promise of retaliation in case of aggression from either side would.
And "we" here means the democratic world - solidarity is the only way to deter bullies - if Russia knew that EU/US would retaliate in case of an attack on Ukraine, they wouldn't invade.
But of course if you yourself is a bully, you will do everything in your power to cause feuds. And here we are, oh well.
In more practical terms, if I were to approach this problem, I'd discretize it in time and apply classical ml to predict "chance to die during month X assuming you survived that long" and fit it to data - that'd be much easier to spot errors and potential issues with your data.
I'd go for the stochastic calculus or actual survival analysis only if you wanted to prove/draw a connection between some pre-existing mathematical properly such as memory-less-ness and a physical/biological properly of a system such as behavior of certain proteins (that'd be insanely cool, but rather hard, esp if data is limited). In my (very vague) understanding, that's what finance papers that use stochastic analysis do - they make a mathematical assumption about some universal mathematical properly of a system (if markets were always near optimal with probability of deviation decaying as XYZ, the world economy would react this way to these things), and then prove that it actually fits the data.
Yes, thunks with sugar is my thinking as well. Does not explain the "last unassignment" examples though. I don't quite understand why it's needed though :)
"This function calculates the minimum of a list of numbers, and subtracts every number in the list by that minimum. This function currently iterates over the list twice; once to calculate the minimum, and once to do the subtraction. At a first glance, it looks like there is no way to merge these two loops, because you need to calculate the minimum of all the numbers before you can subtract from any of them. As it turns out, this is possible, using the inverse operator."
I was long curious about interaction nets. Can someone confirm if Vine will automagically derive an algorithm for doing this with a single pass or it will do one pass in "forward mode", but once the minimum is resolved, follow it up with a second implicit "book keeping and variable re-resolution pass" that will be equivalent to an actual full second pass you'd get in an imperative language?
It is all very much reminiscent of programming in tensorflow 0.x.
I suppose it means per speaker. And it is based on a simplified style tts 2 which from my small dive into the subject seems one of the smaller models achieving great quality.
Differential meshes they show there remind me of holomorphic functions like [1] - is there a connection between such generative processes and minimal surface / curvature?
Yep, that's the exact issue I wanted to address with my own twist on the idea of an online boardgame engine - I was trying to actually persist the callstack of an async wasm vm function (game loop) execution into a database in rust. It is working in a sense that you can implement battle ships or tick tack toe, but I did not quite finish it. Happy to still make repo public if helpful.
To be more specific, async wasm function was implemented as a poll loop sync function exported to the caller, there was (at the time) no way to move wasm mv memory, so it was persisted while the game was live and replayed from a message log stored in a db after/if preemption.