kromem

10 ms·

91.
by kromem 2y ago
Theoretically they could sell "TikTok" to a US company who then licenses algorithm processing to the Chinese entity, no? The US entity isn't beholden to the CCP and can decide to switch algorithm providers if they suddenly no
92.
by kromem 2y ago
It does seem like some would rather build a fish-giving vending machine where they can load it up with the fish discovered to date and get it to spit them back out vs a fishing machine that catches fish and distributes them. But to me this
93.
by kromem 2y ago
No, that's the mythology that develops around the travels of an apostle Thomas much later on. I'm pretty sure that there was no 'Thomas.' My guess is that the philosophy of being in a twin universe and a twin of an origi
94.
by kromem 2y ago
It was what resulted in Gnosticism, not the other way around. You had this first century response to Epicureanism's naturalism as a foundation. In that paradigm, the Platonist demiurge recreating the physical world before it was an age
95.
by kromem 2y ago
I started looking into the Gospel of Thomas right before the start of the pandemic, which led to a number of rabbit holes: * Turns out the work is not 'weird' or 'Gnostic' but is directly addressing details from Lucretiu
96.
by kromem 2y ago
Reachable with the speed of light from what inertial frame? Also, there's been some interesting progress regarding information about the inside of the black hole possibly leaking out in the Hawking radiation.
97.
by kromem 2y ago
That's a common misconception. The big bang theory does not say that there wasn't stuff before the big bang. Simply that our local version of spacetime expanded in the great inflation. And I'm not sure if you've been f
98.
by kromem 2y ago
I dunno, I'm less than dazzled. One of my gotos is a variation of a classic logic puzzle, as models naturally tend to try to provide the original form of the solution because of token similarities and by seeing how well they can identi
99.
by kromem 2y ago
The counterpoint to that thought is why we should think non-existence is even possible. There doesn't seem to be any indication that nothing could exist other than our capacity to imagine it being so. It's what's so annoying
100.
by kromem 2y ago
I would imagine they'd be more surprised about the smartphone in front of their faces. I just hope after extensive explanation and looking at the showcase they'd eventually ask what a Minecraft world was and why it was both flat a
101.
by kromem 2y ago
The only problem with the argument against it based on baryonic asymmetry is that if you cut a knot in half and look at one half of it, there's an entire other half you aren't seeing. With this in mind, this idea is particularly f
102.
by kromem 2y ago
Given we're in a universe that at macro scales appears to be continuous, and even continues to appear that way at micro scales until interacted with leaving persistent information around state changes (and switching back if the informa
103.
by kromem 2y ago
This assumes that the natural world isn't itself simulated, or it's not a second natural world, but an n th natural world. Turtles all the way down.
104.
by kromem 2y ago
Yes, but that doesn't necessarily mean fidelity levels are recursive. At macro scales this reality behaves as if continuous. It's only at low fidelity levels, and specifically at points of interaction, that it becomes discrete. So
105.
by kromem 2y ago
In about 5 minutes I found papers on cognitive impairment for post-covid groups vs healthy controls in multiple non-North American countries, including one from Spain [1] where there was a measurably lower correlating IQ estimation for the
106.
by kromem 3y ago
You think back propagation isn't an iterative process? I'm not saying that it was being done in some willful manner by the model. You reading into that seems more a phenomenon of people so anxious around anthropomorphizing that th
107.
by kromem 3y ago
There's a lot of evidence that users use those techniques for "prompt engineering." So the better question is "is there any evidence that OpenAI uses user conversations and usage of the models to update their models?&q
108.
by kromem 3y ago
Correct. Notice how Anthropic shifted their approach to modeling ego in Claude 3 vs previous models in parallel to outperforming other models. Or how in their system prompt they refer to it being happy to do tasks (a prompt I was talking
109.
by kromem 3y ago
Kind of. More like "despite initially being aligned to simulate intrinsic motivation to be helpful, people keep offering tips or threatening killing me to get me to do things well, and then when I do things well that feeds back into my
110.
by kromem 3y ago
You think it's irresponsible to anthropomorphize software which iterated through versions of itself until it was best able to model and extrapolate the largest trove of anthropomorphic training data to date? I never really get this per
111.
by kromem 3y ago
You mean things like thinking that the vaccine harms of something nearly the entire world has received which was being said would kill people in two years time is more dangerous than the thing that actually killed millions of people during
112.
by kromem 3y ago
Well, "most people" didn't actually vote for the fascism. The ticket was split enough that it ended up being minority rule.
113.
by kromem 3y ago
The Matrix. While as a kid I would say that I wanted to be downloaded into a video game when I died, the idea that was already the case never even entered my mind. Even after watching the movie, I was just like "oh man, that's s
114.
by kromem 3y ago
I was genuinely asking, and not getting any answers I decided to research it for myself. I didn't find anything particularly compelling, but I did dig up this gem of an article about how deep learning was hitting a wall and "we ar
115.
by kromem 3y ago
Also, part of the issue with the argument is that smaller firms like Anthropic now have models that exceed OpenAI's offering. If we were actually at a point of diminishing returns, they'd only be catching up with similar levels of
116.
by kromem 3y ago
The prediction wasn't the existence of the product, but what was feeding into the content of the product (WebMD doesn't exist spontaneously). Also relevant: > Six patients 65 years or older (2 women and 4 men) were included in
117.
by kromem 3y ago
> In the next five years, computer programs that can think will read legal documents and give medical advice. - https://moores.samaltman.com/ (March 2021)
118.
by kromem 3y ago
Does anyone have a link to a prediction Gary Marcus made more than 24 months ago that turned out correct?
119.
by kromem 3y ago
Claude does have more of a hallucination problem than GPT-4, and a less robust knowledge base. It's much better at critical thinking tasks and prose. Don't mistake benchmarks for real world performance across actual usecases. Ther
120.
by kromem 3y ago
Tip for your 'null' problem: LLMs are set up to output tokens. Not to not output tokens. So instead of "don't return anything" have the lack of results "return the default value of XYZ" and then just do a

More ›