macrolime

8 ms·

61.
by macrolime 2y ago
Pretty sure that once the pre-trained models used for this are pre-trained on lots of video and get larger (2-3 orders of magnitude larger than this), things will quickly improve. This may already exist in prototype form behind closed doors
62.
by macrolime 2y ago
How relevant is this for plastic food packaging?
63.
by macrolime 2y ago
I took some AI courses around the same time as the author, and I remember the professors were actually big proponents of neural nets, but they believed the missing piece was some new genius learning algorithm rather than just scaling up wit
64.
by macrolime 2y ago
The VO2Max seems wildly inaccurate. It never activates when I exercise, but sometimes activates when I'm sleeping.
65.
by macrolime 2y ago
Could be the reason is aging is programmed and when food is scarce the program changes to delay aging, such that the organism has a better chance to survive longer and reproduce once food is less scarce and at that point turn up the rate of
66.
by macrolime 2y ago
Not exactly what I would call decentralized training. More like distributed through multiple data centers. Decentralized training would be when you can use consumer GPUs, but that's not likely to work with backpropagation directly, but
67.
by macrolime 2y ago
Does this mean we're closer to getting GPU support on Docker on Apple devices?
68.
by macrolime 2y ago
I think the fact this made it to the front page shows how many were unaware of the scanner in the files app. I have multiple times looked into ways to scan on iPhone and have only found the Notes app, which is kinda klunky, and Microsoft Le
69.
by macrolime 2y ago
Because they want people to buy and use Apple monitors, not third party ones.
70.
by macrolime 2y ago
Brilliant doesn't have spaced repetition, it has much less coverage and doesn't have enough exercises to let you really master subject, but it's an easy to use mobile app with nice visualizations that can give a conceptual un
71.
by macrolime 2y ago
It's the best way to self-study math really. I was previsouly trying using Susan Rigetti's "So You Want to Study Mathematics..." guide. It has been posted here a few times, but it's so easy to get lost in doing tons
72.
by macrolime 2y ago
This kind of problems is really not where LLMs shine. Where you save loads of time is when you need to write lots of code using unfamiliar APIs. Especially when it's APIs you won't work with a lot and spending loads of time learni
73.
by macrolime 2y ago
Source available means you can see the source, but not modify it. This is kinda the opposite, you can modify the model, but you don't see all the details of its creation.
74.
by macrolime 2y ago
For the downvoters, yes a slight exaggeration. The year I checked, 2019 was 1 vs around 30, so just slightly over an order of magnitude if you want.
75.
by macrolime 2y ago
I live in a similarly sized city to San Francisco in Northern Europe and San Francisco has almost two orders of magnitude more traffic deaths each year.
76.
by macrolime 2y ago
If you're afraid of it getting stolen you can buy insurance.
77.
by macrolime 2y ago
With an ebike you can ride in bike lanes, sidewalks and many other places motorcycles are not allowed. They can also park loads of places motorcycles aren't allowed plus you get exercise.
78.
by macrolime 2y ago
I've had complete opposite experiences, there is issues with the derailleurs constantly. Switching to internal gears and a belt instead of a chain has made commuting a much better experience and may in some cases actually cheaper in th
79.
by macrolime 2y ago
Seems like this might be used as an alternative to embeddings for creating semantic search. Run a document collection through this to get the activations, run the prompt through the same thing and use the sparse autoencoder to identify whic
80.
by macrolime 2y ago
I guess because they mostly use Macs. They always use Macbooks in videos.
81.
by macrolime 2y ago
How much energy was spent on the training run though? It's all the energy spent through the lifetime of trillions of organisms.
82.
by macrolime 2y ago
It's not really all that useful for AI training. The reason why energy is a problem for AI training is that you need a lot of compute at one place, distributing backpropagation doesn't work that well. When the energy problems hit,
83.
by macrolime 2y ago
What you perceive is a kind of virtual reality similar to a video game. It's just constructed based on sensory input, but still has a coordinate system and a point of view, like the camera location in a game engine. Electrically stimul
84.
by macrolime 2y ago
Looks like they also have a robot cat https://www.petoi.com/products/petoi-nybble-robot-cat
85.
by macrolime 2y ago
Multisig by the user and camera manufacturer can help to some extent.
86.
by macrolime 2y ago
I wonder then, whether the attacker has already started over with a new open source project, or found something else to do. Or was xz only one of multiple targets where the overall plan didn't actually fail and an updated backdoor, wit
87.
by macrolime 2y ago
Does it support Apple Silicon (accelerated)?
88.
by macrolime 2y ago
Another thing I have noticed is that if you use ChatGPT and it at some points uses Bing to look up something, it becomes super lazy afterwards, going from page long responses on average to a single paragraph.
89.
by macrolime 3y ago
https://www.unciphered.com/blog/randstorm-you-cant-patch-a-h... Over the last 22 months, Unciphered has been working on a vulnerability which affected BitcoinJS, a popular package for the browser based generation of cr
90.
by macrolime 3y ago
"This is more computationally efficient than performing a full content-based lookup across an entire memory buffer for each step in the future, and could be one step towards drastically increasing the context-length available for makin

More ›