Im curious what specifically you think is going wrong. Im a Nintendo fan from the consumer side but dont follow their business, aside from being a huge fan of Iwata’s life and story.
(Im on mobile- sorry for the brevity/lack of deep demo context, it seems like the demo links dont work super gracefully on mobile)
I clicked into the demo (reference 3 in your post) and I dont get how this is different from any off the shelf OTel solution like Lightstep or Jaeger. Otel already has client side vendors and SDKs (I was the engineer who introduced OpenTracing to Plaid, and we did both). Bridging cross-network traces is literally the point of OTel, so Im not sure why thats a differentiator for you… every system does it.
Yep, Ive found that, generally, caching logic should be a separate flow from business logic. E.g wrap your RPC logic such that an in-memory cache transparently sits in your RPC stack. This lets you implement locking on the cache in a single place, and makes it easy to do things like eager prefetching or offloading cache logic to another thread.
Everyone is piling on you but Id love to see what their companies are doing. Cosine similarity and loading a few thousand rows sounds trivial but most of the enterprise/b2b chat/copilot apps have a relatively small amount of data whose embeddings can fit in RAM. Combine that with natural sharding by customer ID and it turns out vector DBs are much more niche than an RDBMS. I suspect most people reaching for them haven’t done the calculus :/
This is really cool - based on the project page it uses off the shelf models like Yolo and Tesseract plus a ton of synthetic data (which is also freely available). Also solves a real industrial use case!
This is true but I think misses the spirit of the criticism
Imagine an alternative world where health insurance acts more like a mortgage with an amortization schedule. when youre young (and have low savings) your premiums act like they do now - simple insurance. But as you age, your payments are split gradually between pure premium and an HSA. The amount you’re personally liable for increases as you age, but so does your ability to pay out of pocket. For high earners this scheme makes a lot more sense (because youll have a few hundred thousand in health savings by retirement).
Most people actually do the above but clumsily and also in the opposite order - HDHP plans are more broadly useful when youre young and healthy because you probably wont have high medical costs so you just save on premiums.
I love radix & shadcn! it's super popular for side projects and young startups since the defaults take you very far, and are very easy to customize. everyone I've introduced them to loves it. It's harder to work into existing projects / design systems at larger co's, though
This is a great characterization of why peer review is great - it makes honest scientists better. A lot of progress is driven by "jumps" from influential papers, and we want those to be as good as they can. It may not stop frauds, but the frauds weren't going to help us anyway. I think fraudulent research mostly hurts by distracting honest scientists and new scientists, not by convincing them of something untrue.
Yea, I think the difference between “how many rotations” and “how far is the center of the small disk traveling” is pretty big. I was completely lost as to how you “add one” until I watched what the video called a “rotation”. It would have never occurred to me to call the coin being upright a rotation… the Art of Problem solving question in another comment thread was more clearly worded for that scenario
This reminds me of a time back in high school when a friend bought a cell jammer on ebay and we walked around school just waiting for people to start texting before turning it on and losing our heads at their confusion. Stupid to do, but bored high schoolers do stupid things…
If its a fund that has led rounds before then yes, but if its a new fund or small fund then it might be part of the GPs proving they can get deal flow to their LPs before raising a larger fund
Ive talked to a few people formerly at OpenAI a few years ago - the deviation from the original mission was a “boiling frog” process that is best demarcated when the Anthropic founders left. The core team doing the actual science and engineering work did legitimately believe in the open research mission. But when funding was hard to find, things kind of broke.
Ya I think most just would not notice between friends. But we’ve compared our iMessages and we aren’t doing anything weird like sharing apple IDs. Not a huge deal so it’s whatever, but if any Apple engineers read this: plz fix
Youre being downvoted but I also lose a message every few months between my girlfriend and I. We both have relatively new iPhones and exclusively use iMessage so I do think its some rare protocol bug.
Yea, this turned out to be the simplest explanation… try to undermine board, get fired, win anyway (lol)
Lesson here is… the next time a startup CEO gets fired they should try the social media spam gambit…? Depends on how much $ is on the line but worked at least once!
I dont think OpenAI is ever going to ahead in image generation, they were lapped very soon after dall-e and every real workflow Ive seen uses Midjourney or Stable Diffusion. The reverse (GPT 4 vision) is well ahead of open source though
This makes sense, but I find it hard to believe the big cloud players won’t have the datacenter skills to compete…
I agree that supply is an issue, but paradoxically the fact that these GPU cloud providers (CoreWeave et al) are partnering with the big cloud players says that the big cloud players are where people would prefer to buy. Once supply constraints are solved, these providers would need some novel offering beyond “we have hardware”, e.g. some specialized distributed training framework. But MS/Google/AWS are also building their frameworks so…
And then the elephant in the room is: compute spend so imbalanced on training vs inference. Why? Is it that there arent enough real use cases? Is it that improvements are so frequent it makes sense to toss out older versions? Is it that privately trained models are a requirement for the highest spenders? My impression is that a lot of corporate spend at scaleups is purely speculative r&d to evaluate capabilities but thats a small sample from friends
Eh, this is exploring a more “static link” approach for local use and development vs the more common “dynamic link” that API providers offer. (Imperfect analogy since this is literally like a DLL but… whatever). Probably makes sense for private local apps like a PDF chatter.
The models dont look bad! But based on my dives into gaming, the modeling work for animating anything that moves (rigging, weight painting, retopology, creating animations, lighting) is still unsolved. This is great for static assets like boxes or things that add flavor to an environment.
For Amazon, it was actually about vastly expanded inventory. It was often hard to find books if you didn't live near a large retail book chain. I remember having trouble in high school finding some english class books driving around the suburbs.
well, this assumes the chat (where the ratings are given) is what people are using and paying for. I think most businesses pay for some combination of API access and specific use cases like code generation (at least, thats what I pay for) that don't really impact RLHF data. General search for consumers is likely to schism since chatGPT isn't especially different from Bard or Edge's AI assistant or the myriad of other product surface areas that can add it.
I use Bard simply because its one less login screen (ChatGPT has an aggressive cookie expiry afaict) and it has up to date info with sources typically available. If it could reliably provide source info thatd be great for API/library documentation while programming or random numerical and finance (what is X company’s price to sales?) searches I do all day.
This all depends on product requirements. But the single simplest reason for choosing a cloud provider like AWS is reliability. If a region or machine goes down, how do you handle it? AWS has the best reliability among cloud providers, so by choosing them, the odds you even need to think about it are minimized. And they have answers to multi-AZ and regions. But if a few mins of downtime every few months isn’t the end if the world, by all means save yourself the trouble until it is.