solarwindy

8 ms·

31.
by solarwindy 1y ago
I think you neatly summarise why the current pre-trained LLM paradigm is a dead end. If these models were really capable of artificial reasoning and learning , they wouldn’t need more training data at all. If they could learn like a human
32.
by solarwindy 1y ago
Very much appreciate this philosophy to nutrition and physical development. Why supplement vitamin B12, incidentally?
33.
by solarwindy 1y ago
Then again... Our brains only developed their abstract reasoning capabilities after we already possessed a cognitive ‘platform’ for learning in a very physical context. Or rather, our cognitive capabilities for things that may seem totall
34.
by solarwindy 1y ago
Hmm, I should say that in the legacy case, it’s only been with legacy I’m already (painfully) familiar with, and knowing how difficult that code has been before to disentangle or to carefully stack another card on the already teetering hous
35.
by solarwindy 1y ago
My comment was admittedly a little inflammatory, so I’ll try to elaborate on what I find so exasperating about the whole experience of using LLMs for coding (in any form, chat, agent, etc.): at root, it’s their total lack of a model of comp
36.
by solarwindy 1y ago
Juniors do learn though, by contrast (as you said yourself). So these models are, at best, perpetual juniors . Except, they make bizarre mistakes that human juniors would not, and which are hard for even experienced developers to spot, b
37.
by solarwindy 1y ago
Having intermittently given in to the hype and tried whatever is supposedly latest and greatest, it without fail goes like this. I’ve even on occasion deceived myself into thinking I’ve finally hit on the right way to use these tools, only
38.
by solarwindy 1y ago
All too often with AI proponents I get the you’re holding it wrong vibe. I can’t help but develop the suspicion these people are putting out bad work and just never notice. Or to be a little less scathing, fine, there’s a very narrow ra
39.
by solarwindy 1y ago
Granted. Although, setting aside the question of sentience, there’s a more serious point I’d make about the dissimilarity between the always-on nature of human cognition, versus the episodic activation of an LLM in next-token prediction—nam
40.
by solarwindy 1y ago
The phenomenon of waking up before an especially important alarm speaks against the notion that our cognition ‘stops’ in anything like the same way that an LLM is stopped when not actively predicting the next tokens in an output stream.
41.
by solarwindy 1y ago
Agreed, though long before such qualms come to the fore I'd like to see even a shred of evidence that this entire approach to AI is at all capable of formulating mental models of the kind that have enabled humans to produce all the won
42.
by solarwindy 1y ago
The managers may believe that's what they're asking their developers to do, but doesn't this whole charade expose the fact that this technology just does not have even close to the claimed capabilities? I see it as wishful th
43.
by solarwindy 1y ago
> On two occasions I have been asked, ’Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?’ I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
44.
by solarwindy 1y ago
Your comment immediately brings to mind this classic of croissant eating in ‘interesting’ circumstances: https://youtube.com/watch?v=wp84sRpM1Js
45.
by solarwindy 2y ago
That supposes gen AI meaningfully increases productivity. Perhaps this is one way we find out.
46.
by solarwindy 2y ago
Not quite fully expendable. Commentary during the livestream was that they’ll recover the fairing, for what it’s worth.
47.
by solarwindy 2y ago
Fun thing about Friday, there was no moon to see by. But absolutely, the full moon on a clear night can be a floodlight, especially with snow on a nighttime ski tour.
48.
by solarwindy 2y ago
Last night in particular, hiking through the forest there was no need for a flashlight at all.
49.
by solarwindy 2y ago
Pretty sure Lunar [0] can do this for you, and you can buy a lifetime license. [0]: https://lunar.fyi/
50.
by solarwindy 3y ago
They prefer the latter because the mere existence of the knowledgable, informed, involved engineer questions the need for the manager at all.
51.
by solarwindy 3y ago
And yet, the specification is not what executes. If the code can be written by a domain expert, why compromise? If nothing else, more time spent working on the problem by a domain expert gives more opportunity to identify deficiencies in
52.
by solarwindy 3y ago
Any high signal/noise subreddits you can recommend?
53.
by solarwindy 3y ago
Anyone in particular you recommend to follow?
54.
by solarwindy 3y ago
Now you have me pining for the sansho pepper ramen I will never forget in Ginza.
55.
by solarwindy 3y ago
Ah, so if the pollution is widespread enough, it’s impossible to trace the blood-borne PFAS of the plaintiff directly to the defendants. Good job PFAS is everywhere now.
56.
by solarwindy 3y ago
Here’s a few artists you might like: @FEELSxart @ilithya_rocks @generativelight @thresfold @Tezumies @_nonfigurativ_ @Olga_f2727