BosunoB

Honestly, open weight models are dangerous, and I haven't seen a single convincing argument against that.

The specific method outlined here seems pretty doable by today's models, which don't seem smart enough to have some kind of 4d chess hidden alignment.

You just have to go through a ton of all of a model's neurons and identify the associated motif, then use that model as a research model to figure out how to tune out bad behavior.

It's pretty hard to imagine today's models subtly recognizing a motif for bad behavior in the research model and then lying about it. Or lying about the results of an alignment training run. It's possible, I guess, but it seems unlikely, and I don't really think there's any evidence of current models having that variety of misalignment.

If the models get good enough, the demand for them will soar.

These companies believe they're a year from AGI, and that's how they'll be profitable.

Go ahead and ask it for music that's similar to a song you like. For that use case, I promise it's just way better than anything in the past. Pandora, Spotify Radio, anything.

In my experience, I could choose like, Steve Lacy radio from the song list, and it won't give me anything that sounds like Steve Lacy. It'll give me music that people who like Steve Lacy also like.

But, if I ask the DJ for music like Steve Lacy, I'll get music like Steve Lacy.

I think the LLM basically has a much better semantic understanding of music and how it sounds than the traditional like-based algorithms.

I do agree with you on the base DJ, it just jumps between homogenous blocks, but the impressive thing is asking the DJ for a certain kind of music.

Spotify seems to use an LLM for their new DJ feature, and it honestly is way better than any other music recommendation service I've ever tried. It will give you music that actually sounds like the music you ask for.

What I'm excited for is Bumble's AI matchmaker. LLM categorization may actually be the key to a decent dating app.

I think it more likely reflects using bigger phones, where swiping across a while keyboard is quite slow, but also, I hate typing with two hands with any other app than Fleksy. You just have to hit within the general area of the key and it counts. I haven't had that kind of success with any other keyboard I've tried.

The thing is, with Fleksy, as long as you get within a few characters of the letter, it counts, and you also switch auto-correct choices by swiping up or swiping down, so you can fix them easily. It's a joy to type with 2 hands, honestly.

But none of us decide what our cage looks like, now. Our cage now is capitalism, wage labor, disease, death, suffering. We already are tossing the environment in the meat grinder, and many people already have very limited freedom of movement due to poverty, government choices, or circumstance.

You can argue that, at least for democratic countries, these are choices we've made of our own self-determination, but is that even fair to the millions or billions who have far less self-determination?

The cage imposed by nature or by the leaders of whatever country you're born in is just not a very good cage for a lot of people.

Is a human zoo such a bad outcome? We stake our identities on being in control and on top, but do we need to be? Clearly the author is invested in his image of us as rulers—this is an emotional appeal as much as a pragmatic one.

Say we solve the pragmatic argument of our extinction, and our AIs view us, as Geoffrey Hinton likes to say, how a mother views her baby? Then we give up our control to them, and in return we are given lives of leisure, luxury, with minimal suffering, the end of disease, etc. Is this such a horrible scenario?

I don't really get swiping. It's always going to be quicker and more efficient to type because you can use both hands, right? And you don't have to swipe your finger across the screen?

Still going strong with Fleksy even though you can't even download it anymore.

13 billion in revenue, 7.5 in cost of revenue. Can we finally put to bed the idea that inference is subsidized?

Fable was trying to verify a UI change in my game. I was working in another window and noticed a program opening on my task bar. Fable had opened the game through the CLI using a movie maker tool, recorded the output, took a frame from the end of it, and used that to verify the UI. When my game's welcome screen obstructed what it wanted to see, it created a temporary worktree, deleted the welcome screen, and ran the movie maker again.

I watched the whole thing thinking it could've just asked me for a screenshot and saved the tokens. But still, I couldn't help but be impressed. Opus never would've done that.

Very well said. People are making a lot of claims when very little knowledge of the financials is public. If you actually look at the numbers, there are plenty of ways in which API revenue and forgotten subs could more than make the difference for power users. Even if power users are getting 10-20x their sub fee in tokens, the math could still work out. Personally, I doubt more than 5% of Claude subs even approach max usage, because it requires having so many agents running all of the time.

I imagine we'll know in a few months when these companies go public.

They are losing money because they are training new models and building new data centers. The claim of the video is that they're losing money just serving current AI models. There's just no evidence of that.