hedgehog

32 ms·

631.
by hedgehog 3y ago
Think of this like incremental builds via make etc. You don't necessarily trust the output 100% but it speeds up work so much it's worth it, you can always do a clean build at key points where you want more confidence.
632.
by hedgehog 3y ago
This is probably not exactly right but here's a way to think about it. Look at the diagram on the Wikipedia page. Set aside parasitic drag for a moment, and imagine the loop as four segments (up, downwind, down, upwind). Imagine the gl
633.
by hedgehog 3y ago
For anyone who hasn't seen the video: it's really worth watching with headphones on. It looks absurd and the sound is familiar to anyone who has spent any time around fast jets, but it's a glider.
634.
by hedgehog 3y ago
In my opinion the main issues are a layer up from that, it's more of a customer experience and brand building exercise to solve by actually improving the tools people use. Today NVIDIA is the default if you want to get work done and no
635.
by hedgehog 3y ago
Hey! Yep, this is Choong.
636.
by hedgehog 3y ago
Across the semi industry the prevailing attitude is that the chip is the important part and software is a nice add-on that you build as a one-off or that "the community" will support. This has changed much more slowly than I thoug
637.
by hedgehog 3y ago
Having the relatively unique experience of trying to sell an alternative deep learning stack, and then selling that company to Intel, color me skeptical. AMA.
638.
by hedgehog 3y ago
Exactly, nice work BTW. And no hate for Mistral, they're doing great work, but let's not confuse weights-available with fully open models.
639.
by hedgehog 3y ago
I'm well aware of the many open source architectures, and the point stands. Models like GPT-J have open code and data, and that allows using them as a baseline for architecture experiments in a way that Mistral's models can't
640.
by hedgehog 3y ago
It's a big distinction, if I want to tinker with the model architecture I essentially can't because the training pipeline is not public.
641.
by hedgehog 3y ago
If you believe a few assumptions then foundation model companies like Mistral look very valuable even if they open their models: 1) few companies are be able to produce top performing foundation models at any given time, 2) this will contin
642.
by hedgehog 3y ago
Looks like Apple figured out a way to identify Beeper clients: https://techcrunch.com/2023/12/08/apple-cuts-off-beeper-mini...
643.
by hedgehog 3y ago
derefr cited one reason but there's another that's relevant to this thread: updates. In the Android model handset manufacturers and carriers decide when (or if) to ship updates. Google distributing their apps through the store giv
644.
by hedgehog 3y ago
Who cares what the color is? One message type is E2E encrypted over Apple's infrastructure, and you can message while not on cell. The other is not E2E and may have carrier billing and whatever other restrictions on it.
645.
by hedgehog 3y ago
Apple maintains iMessage compatibility with devices that are long out of support, if Beeper Mini is sufficiently similar to the client in for example iOS 12 then it makes an Apple decision to break Beeper fairly expensive. Even if they do t
646.
by hedgehog 3y ago
The binder I wish I had would have schematics for the structure and systems, maintenance records, checklists, vendor receipts, and manuals. No such luck.
647.
by hedgehog 3y ago
Most of the images there are broken, all I see are some conceptual block diagram type things. Find a teardown of the hybrid unit and compare to a teardown of any automatic, look at the actual internal complexity and moving parts. I was surp
648.
by hedgehog 3y ago
I haven't looked at the data but I'd expect the automatic emergency braking is helping reduce accident severity significantly as well.
649.
by hedgehog 3y ago
Car Play has been reliable for me in a Toyota. Now the Toyota software, twice in two years I've pulled over to power cycle the car because the head unit was wedged. I'm not sure how that compares to other cars but never had to reb
650.
by hedgehog 3y ago
I spent some time looking at this and it depends which hybrid, there is quite a bit of variation in drivetrains even within manufacturers. As sibling comment [1] describes the planetary gear arrangement in most Toyotas is quite simple compa
651.
by hedgehog 3y ago
This is not true, at least for the most popular Toyota hybrid drive train. It's probably got fewer moving parts than the ICE counterpart by virtue of dispensing with the automatic transmission. See sibling comment: https://n
652.
by hedgehog 3y ago
I tried switching to Duck Duck Go a while back and so far haven't returned to Google. Some things are better on Google but for my usage pattern it doesn't seem to be a clear win any more, and I'm somewhat picky. What I take f
653.
by hedgehog 3y ago
I imagine they'll replace the physical cards as the existing ones have Goldman's name on the back. The transition could otherwise be nearly invisible, the Goldman name and actual card numbers don't show up often in normal usa
654.
by hedgehog 3y ago
The parent post suggests that Google's moat is search quality, I suspect that is not true. Google earns about $70-80B a year, if they even suspected there was a way to reduce or eliminate a $20B yearly expense it would be top priority
655.
by hedgehog 3y ago
If Google really has a moat why do they pay Apple so much to be the default search engine and why do they need Android to avoid getting cut out of mobile advertising?
656.
by hedgehog 3y ago
Another thing to note here is this looks to be around seven total days of training on at most 4 A100s. Not all really cutting edge work requires a data center sized cluster.
657.
by hedgehog 3y ago
This looks really cool, and it's good to see Keras come full circle back to being an ergonomic interface with support for multiple frameworks. I remember being pretty disappointed when pluggable back ends went away. I appreciate that i
658.
by hedgehog 3y ago
Highly relevant if you want to work on ML systems. Despite how much OpenAI dominates the press there are actually many, many teams building useful and interesting things.
659.
by hedgehog 3y ago
Aha, that would do it. As an aside the amount of stuff baked into these models is impressive, I tried using the dolphin2.2-mistral 7B model to translate bits of that ticket and it did just fine. >>> Translate: 如题 The title su
660.
by hedgehog 3y ago
I think something is off about the current Ollama support, at least with the smaller 6b model it seems to generate near gibberish.

More ›