randcraw

13 ms·

181.
by randcraw 2y ago
Is this new? Or has this been true for decades? What is the trend, both across countries and within each?
182.
by randcraw 2y ago
True, but I think any form of supervised ML can and often does suffer the same fate. Once you define any loss function, you narrow the signal to a subset that presumes the signal and its distribution are understood and reliable. That'
183.
by randcraw 2y ago
A similar article at The Atlantic from September by Olga Khazan: https://www.theatlantic.com/ideas/archive/2024/09/loneliness... I think both articles make some good points, esp outside the topic of lone
184.
by randcraw 2y ago
Thanks. Now, after almost two years of incomparably explosive growth in LLMs since that paper, it's remarkable to realize that we still don't know if Scarecrow has a brain. Or if he'll forever remain just a song and dance m
185.
by randcraw 2y ago
Respectfully, if you're driving an Insight, you probably never pass anyone.
186.
by randcraw 2y ago
Exactly. If your objective is to control your weight, changing your diet has probably 5x to 10x more impact than changing your physical activity.
187.
by randcraw 2y ago
My 2019 Honda Ridgeline has slammed on its brakes at over 60 MPH half a dozen times when it decides I'm passing another vehicle too closely. I HATE the damned thing. I'd turn it off, but it resets back on each time I restart the
188.
by randcraw 2y ago
I suspect that at age 8, merely draping a blanket over a coffee table (or a riding a magic dragon named Puff) is no longer as convincing a fantasy of Camelot as it was at age 3. Such fancies don't necessarily die off at some sell-by a
189.
by randcraw 2y ago
I wonder which will benefit more: the reversal of aging or the scar-free healing of wounds?
190.
by randcraw 2y ago
Presuming human and LLM error to be equivalent assumes that the risk of committing errors by an LLM has the same distribution that human errors do. But they don't. LLMs make thunderously insane errors in ways that no human would do -
191.
by randcraw 2y ago
Can't you do the same thing by pouring half a capful of H2O2 into your ear while lying on your side? After it stops fizzing (about 10 minutes) you sit up and catch the effluent with a tissue. And no need to add oil or carefully shoot
192.
by randcraw 2y ago
Each medical innovation overcomes but one hurdle on a path to a treatment/cure, often revealing new hurdles that weren't visible (or important) until the first hurdle fell. A good example is lipid nanoparticles, which for the firs
193.
by randcraw 2y ago
So the new Mini is faster than all existing Macs (at least on multicore Geekbench). That revelation sort of throws a monkey wrench into the grand plan of seldom buying a super-duper Mac that will perform competitively for years. Better to
194.
by randcraw 2y ago
You don't think SEO-LLMs will evolve to redirect search-LLMs to 'see the world' the way the SEO-LLMs want it to? I foresee SEO-LLM-brinkmanship as the inevitable outcome. Soon THIS will be the catalyst for the real Skynet -
195.
by randcraw 2y ago
I really don't see the value of summarizing/repackaging web search hits. Given that 99% of SEO-tuned web content is just shilling for vendors who don't want to be seen, LLM search summarization will just repackage those ads
196.
by randcraw 2y ago
https://sites.pitt.edu/~bertsch/Moving%20to%20a%20World%20Be... In case you want your own PDF of the paper.
197.
by randcraw 2y ago
Thanks, but site broken. It prompts with an infinite series of images.
198.
by randcraw 2y ago
How could ANYBODY vote against this?
199.
by randcraw 2y ago
One think I like about this effort is their attempt to factor out the cacheing of prior answers due to having asked a similar question before. Due to the nearly eidetic memoization ability of LLMs, no cognitive benchmark can be meaningful
200.
by randcraw 2y ago
Yeah, that kind of reasoning (abduction) tends to be included with deduction and induction as the third form of reasoning. But IMO, it's really not. It's essentially confabulation, or legalistic argument -- where you're not
201.
by randcraw 2y ago
Interesting theory. The notion that some expressions of personality are obstructed by social 'noise' and thus benefit from solitude makes a lot of sense. By their reckoning, solitude is positive, voluntary and purposeful, the me
202.
by randcraw 2y ago
This essay sort of waves a hand at the sub-symbolic roots of knowledge that lie beneath text and that babies spend several years mastering before they are ever exposed to text. IMHO the proper measure of that latent knowledge is qualitativ
203.
by randcraw 2y ago
Only three can win. Do those two deserve recognition more than Hopfield?
204.
by randcraw 2y ago
IMO this award recognizes the effort and success of modeling the acquisition of knowledge, especially toward realizing such models in usable ways that will surely redefine life on this planet forever. Few Nobels have recognized work that i
205.
by randcraw 2y ago
Not AI = A?
206.
by randcraw 2y ago
It should make him uniquely qualified to discuss whether writing with an LLM can ever be as immersive an experience as writing without one can be.
207.
by randcraw 2y ago
Sorry. Meant Constantine (350 AD) -- far more significant to the rise of Christian morals than Charlemagne.
208.
by randcraw 2y ago
A very interesting article, especially in that historians still aren't sure why war-based slavery died out completely in Britain in only 60 years -- between 1066 and 1120. You have to wonder if there was mostly an economic reason for i
209.
by randcraw 2y ago
NP-hardness was a popular basis for arguments for/against various AI models back around 1990. In 1987, Robert Berwick co-wrote "Computational Complexity and Natural Language" which proposed that NLP models that were NP-hard
210.
by randcraw 2y ago
I'm no expert, but I've been looking into the prospects and mechanisms of automated reasoning using LLMs recently and there's been a lot of work along those lines in the research literature that is pretty interesting, if not

More ›