mvkel

30 ms·

601.
by mvkel 2y ago
It may actually be improved radar tech, so finding smooth air is easier
602.
by mvkel 2y ago
No. Why use "widgets" and "electronic flow meters" as analogies? The average person knows even less about those than mortgages. Here's my attempt: Mortgages are not primarily a service that a bank provides to homeow
603.
by mvkel 2y ago
> If you replaced your current mental model for mortgages with “it’s like a paper electronic flow meter for money, possibly with less paper these days”, it would improve your ability to understand the mortgage industry. I feel even more
604.
by mvkel 2y ago
When you read a website do I get to say you're plagiarizing it?
605.
by mvkel 2y ago
Excellent comment. The lack of nutrition plays. Impossible meat, for example, is extremely processed and generally much worse than red meat. But, it doesn't require killing a cow to acquire. I wonder if this is more appropriate for red
606.
by mvkel 2y ago
Nope! From the grocery store straight to the shelf in the fridge, upside down. It actually mixes fairly well even at room temperature, too
607.
by mvkel 2y ago
This is neat! But you can also just store the jars upside down in the fridge. No need to stir.
608.
by mvkel 2y ago
Great comparison. The comparison itself is pretty telling; that the brain and AI work so similarly in specific ways
609.
by mvkel 2y ago
Until <5 years ago, AI was almost entirely a purely academic field, theoretical at that. Those same academics admit themselves that they're surprised at how well LLMs do considering how simple(?) rudimentary(?) the logic underneath
610.
by mvkel 2y ago
Just because you can make something doesn't mean you know why it's made. There are thousands of people around the world trying to reverse engineer what is going on in the billions or trillions of parameters in an LLM. It's a
611.
by mvkel 2y ago
On what evidence are you making this statement? We don't even know how existing LLMs work, not really, yet we're done? All signs are pointing to there still being no upper limit on results. Not yet. Just because it's not on t
612.
by mvkel 2y ago
The SaaS business model breaks if you go single tenant except in Fortune 500 enterprise
613.
by mvkel 2y ago
How does one technically opt-out after model training is completed? You can't exactly go into the model and "erase" parts of the corpus post-hoc. Like when you send an email to feedback@slack.com with that perfect subject lik
614.
by mvkel 2y ago
Most, if not all SaaS software is multi-tenant, so we've been living in the "will not" world for decades now.
615.
by mvkel 2y ago
Because it's default opt-in, and most people won't see this announcement.
616.
by mvkel 2y ago
The best LLMs were trained on data from the open internet, which is full of garbage. They still do a pretty good job (granted it has been fine tuned and RLHF'd, but you can do that with Slack data too)
617.
by mvkel 2y ago
> we do not build or train these models in such a way that they could learn, memorise, or be able to reproduce some part of Customer Data They don't "build" them this way (whatever that means) but if training data is someh
618.
by mvkel 2y ago
Hm? No, it really did come from a front-end person. There was a period in the late-aughts when people wanted to emulate the iPhone's inertial scrolling on the desktop. Most modern sites had it and it was infuriating. That's probab
619.
by mvkel 2y ago
Because the commenter implied that LLMs get it wrong, as if search engines get it right. The reality is that both take digging, but the LLM response gets you to the right answer more quickly.
620.
by mvkel 2y ago
If you consider the best frontier LLM in the world "not that impressive," I'm curious to know what you think is
621.
by mvkel 2y ago
Because their weakness is in focus and execution, not resources. If they focused all of their energy on, simply, a frontier AI model, and not trying to shoehorn a half-complete model into all of their products, there is no doubt they would
622.
by mvkel 2y ago
Here are the results for that exact search in Google: https://www.google.com/search?q=what+was+the+last+non+cellul... vs 4o: The last non-cellular Windows Mobile Classic PDA was the HP iPAQ 110 Classic. Released in 2008, i
623.
by mvkel 2y ago
GPT-4o serves up results that look like Perplexity's, except the sources are actually relevant links. All of that to say: solved problem? Assuming you're ok with chat as the UI
624.
by mvkel 2y ago
We don't need to. We have LLMs now
625.
by mvkel 2y ago
It's wild that a one-time convenient excuse turned into the base expectation of all social media that is impossible to live up to. People are weird.
626.
by mvkel 2y ago
Catching up right now is not a matter of tech innovation, but raw energy and compute. Of course, the next -revolution- in AI could very well come from Ilya. But why would he bestow that honor to anyone? He can self fund it if he wants. It&#
627.
by mvkel 2y ago
Calling social media a "public square," isn't a valid analogy, and is an impossible ideal to live up to. Nothing about it works the same way an actual public square does. It's a coliseum.
628.
by mvkel 2y ago
It's already here. Demo: https://youtu.be/kkIAeMqASaY?si=gOspEouo69eQmDiA I also have an anecdote where it served (successfully) as a mediator for a couple. Exciting times.
629.
by mvkel 2y ago
I was in the audience at the event. The only parts where it seemed to get snagged was hearing the audience reaction as an interruption. Which honestly makes the demo even better. It showed that hey, this is live. Magic.
630.
by mvkel 2y ago
Of the -really- talented tech folks I know, the opposite happened.

More ›