Fair point. Maybe I should've said SteamVR environments instead, it never took off like SourceSDK did, partially due to incomplete tools. SteamVR as a whole is very healthy though.
S&box was based on UnrealEngine 4 until late 2020. I think Garry wanted to use the latest and greatest engine, then Valve continued to be friendly with him, and even though Valve wanted Source2 to be a VR platform, it was clear desktop was going to remain relevant, but the content creation tools on SteamVR Environments were a cut-down version of what was actually used to make Half-Life Alyx, but they gave Garry all the tools, and he moved to Source2, and built a .net "framework" to make it faster to develop and iterate in Source2. So Garry's tools are now an open version of the closed tools that Valve didn't want to release that they used to make Alyx.
Finally there's another serious competitor to UE and Unity.
I know Veritasium gets posted here a lot, but a few days ago he posted a deep-dive into the the engineering of jet engine turbine blades. Turns out they're made from a single crystal of a superalloy that entangles itself at a molecular level such that it actually gains strength as it's heated, only losing strength above 1200 degrees C / 2200 degrees F. Below that temperature, as long as the strain on the part is below the plastic deformation threshold, it's not really losing any strength at all over time.
LEDs are around 10x more efficient than incandescent bulbs. A 100W incandescent equivalent LED bulb typically consumes 10-12W. A 70W LED would put out as much light as 700W of incandescent.
Sure, but having a tactile bump in the travel makes it that much easier. I can see the argument that it might seem overcomplicated or confusing to typical users though.
Yes, that was my point, Mac Mini solders components that are not soldered on most other computers of that form factor, but a socketed CPU or GPU would be extremely unusual.
There are groups that are more focused on music (DnB, dubstep, other festival-friendly genres), focused on dancing, focused on drinking games, focused on world-hopping, etc. I'm into the underground rave vibe, so for that there's VRC Party Hub, which is a guy who runs a discord who befriends as many clubs as he can find in that scene, and imports their schedules/announcements channels into a nightly report of all known events.
I used colorcontrol to enter the service menu of my LG C2 and disable all anti-burn-in features. No auto dimming, no auto picture level, no anti-logo, etc. The only one I kept is pixel shift because it's only noticeable if you're looking at the edge of the screen when it moves, it's a tiny movement. I skip the "pixel cleaning" prompt every time it wants me to wait. When I'm gaming in HDR, I use filters to increase the exposure to get the maximum brightness range of the panel. Been using it like that for ~8hrs/day for over 2 years now. Zero detectable hint of burn-in.
Not GP, but I imagine because going back and fourth to compare them is a waste of time if Gemini works well enough and ChatGPT keeps going through an identity crisis.
Around the mich med main campus/huron river, in the winter, there will be a couple of days here and there that thousands and thousands of crows decide to blanket the campus with poop and caw at you as you try to navigate it with your feet. Good idea to have a hood up or hat on.
I got the ultra dual drive (non-go), no silly spinny bit, just a sliding mechanism that locks into place in either configuration as well as fully retracted on both sides.
This is all extra confusing (as to why people republished this) because a 9850X3D was already rumored a couple weeks ago as a higher binned 9800X3D, which would actually make sense, as well as a 9950X3D2 with dual X3D CCDs.
Yes, that's before the term got corrupted. If you've spent any time in the depths of brainrot mass social media (or know anyone in bigcorp HR), you'll find that there are some large cohorts out there who have taken this to an extreme that looks exactly like that type of racism but "flipped." The masses are struggling to imagine a more fair society rather than seeking revenge.
Do you have any links on tracing a NN for a hallucination/unconfidence neuron? I do worry that it's possible there isn't an obvious neuron that always goes to 1 when bullshitting, but maybe with the right finetuning, we could induce one?
I am aware, but think about right after a smaller is done training. The researchers can then quiz it to get a sense of the depth of knowledge it can reliably cite, then fine-tune with examples of questions beyond the known depth of the model being refused with "Sorry, I'm a small model and don't have enough info to answer that confidently."
Obviously it's asking for a lot to try to cram more "self awareness" into small models, but I doubt the current state of the art is a hard ceiling.
To be sure, a lot of this can be blamed on using AI studio to ask a small model a factual question. It's the raw LLM output of a highly compressed model, it's not meant to be everyday user facing like the default Gemini models, and doesn't have the same web search and fact checking behind the scenes.
On the other hand, training a small model to hallucinate less would be a significant development. Perhaps with post-training fine-tuning, after getting a sense of what depth of factual knowledge the model has actually absorbed, adding a chunk of training samples with a question that goes beyond the model's fact knowledge limitations, and the model responding "Sorry, I'm a small language model and that question is out of my depth." I know we all hate refusals but surely there's room to improve them.
There should probably be a little more effort towards making small models that don't just make things up when asked a factual question. All of us who have played with small models know there's just not as much room for factual info, they are middle schoolers who just write anything. Completely fabricated references are clearly an ongoing weakness, and easy to validate.