Incurable ignorance is a perfect position for a long passionate arguments.
299 karma · joined April 4, 2024
Previously on HN:
Shannon Got AI This Far. Kolmogorov Shows Where It Stops
https://medium.com/@vishalmisra/shannon-got-ai-this-far-kolm...
...Therefore abstraction and formal methods.
MacOS and apps have many kinds of background activity which present large transient loads, for example Spotlight indexing and photo library facial scanning.
Do you have a browser or other app infected with a crypto-miner or other malware?
If you create a new user and sign in using only that new user, do you see the same overload?
Are you running an unsupported Mac for Tahoe via OCLP? Thar be performance hazards. Search Macrumors "unsupported" forums.
As models are under direct control of totalitarian silos, ideology can be baked into every tendril of knowledge represented.
A service such as Gemini may be a societal ideological risk on a greater scale than the notoriously fascist influence of radio/TV in the 1920s and 30s.
"Guardrails" are direct lines of injection of ideology, and "hallucinations" can serve to mask.
Why should anyone trust singular corporate control points for doctrine on the totality of knowledge? It seems like a crazy risk.
News is already utterly polluted with doctrinaire bias.
Politics are clearly designed to keep the electorate away from decision-making, and open doors to privileged silos of control.
AI can own the populace coming and going as the models inject doctrine into the core of knowledge and scour the consumer for measures doctrinal uptake and alignment. Will information become totalized with no possibility of escape?
Humans can't keep up with the complexity of AI; we're going to have to let AI manage it
How was the port verified as complete?
Having your phone number listed was once a desirable service of the phone company
The cloud GPUs are 170 thousand a year.
1000 minis are going to cost 1.5 million + floor space, power, setup and administration over their lifetime.
Amortizing the minis against the cloud savings looks to require 10 or more years.
But the entire genAI industry has manifest in less than 10 years-- AppleSi based compute modules aren't that old!
—The Grand Turk
Fixed it.
https://www.0xkato.xyz/how-llms-actually-work/
Or are they sourced from a common ancestor?
I think, therefore I am.
Try it yourself.
If that's insufficient for you, you'll have to deal with this on your own.
Carlin's stand up was fun as loved goofing with such ambiguities.
Two utterly unnecessary and pretentious anecdotal figures of speech in this article are as devastating to the vibe as the slop it complains about:
I want to be clear...
...matters more than you realize...
The more woke the point, the more likely these figures will be used.
Closely followed by:
Let's get this straight..."
Did you know that..."
A list ordering is the most primitive and least memorable layout because lists sort arbitrarily and alphabetical listing of capabilities are not intuitive.
But the weirdness only grows from here:
For example, Photos shows library recents bottom to top, but pick-photo from library shows recents top to bottom
Portrait orientation puts "Done" on one end, landscape puts it on the other.
"Done" can be implied by a return tap or involve a "done" tap.
Some controls tap, some slide and some do both.
Release to release, the formats move around.
Format varies between apps & modes.
Mystery meat abounds
Holding the device a certain way causes spastic mode changes, which vary release to release.
Almost any way you touch the device instigates an action or mode change and some controls have 3+ levels of function:
WTF does the "power" button do?
- stand-by - camera shutter - emergency SOS vs shutdown - arbitrary mode change depending on accessibility setting
Bugs and features overlap.
The UI is never baked, ever more modal...
exhausting
Corneal Electromechanical-Reshaping (EMR) experiments suggest that optical correction of vision may someday be possible by reshaping the cornea via electro-chemistry and a corneal mold, rather than tissue removal. Animal testing shows promise.
Fat cats please begin boarding now
The same publication posted 2019 article on Tycho Brahe. This article quotes that article:
"Brahe saw a star" (something-something) "AI!" The end. (profit)
Just read the other article, and forget AI, as the OG intended:
"And monkeys might fly out of my b—"
https://medium.com/@vishalmisra/shannon-got-ai-this-far-kolm...
This article explains what's missing in terms of two kinds of complexity that oppose: Shannon complexity vs. Kolmogorov complexity.
It introduces the opposition by an example of driving the value of pi as decimal number, which has no pattern and high complexity, and a formula for deriving pi that does have a pattern with low complexity, then observing that mind can work from the patternless high-complexity back to the patterned low-complexity without prior examples, while AI can't.
LLMs encode and retrieve patterns in the training data, and doing so can connect data to the terminology of known principle, but mind can observe inconsistencies in data and to reason from first principles to resolve the inconsistency.
The distinction between these two modes can seem blurry as AI can traverse the patterns of the known in ways that are extraordinarily revealing, but it's not structured to reason about the unknown.
Inference is not sufficient for reason.
For example, a conventional algorithm can search for patterns in text at a scale many orders of magnitude beyond a mind's capacity, and this can be very revealing, but to do so this algorithm need not read the text with comprehension.
Regarding the question: can genAI be enhanced to reason? The answer is assumed to be "no", due to the categorical opposition of the two kinds of complexity and the lack of understanding of structures within genAI to handle the reasoning.
Read the article, which includes other examples including a jump from Newtonian to Einstein physics in the history of astronomy, and a noodling on how to talk about the edge of the unknowable in AI.