youssefabdelm

Not at all... they're still so incapable of so much. And even when they do advance, they can be tremendous tools of synthesis and thought at an unparalleled scale.

"A good human plus a machine is the best combination" — Kasparov

Those who seek the easy way do not seek the true way — Dogen Zenji

Hm for me interactive visualizations are more illuminating:

https://www.falstad.com/dotproduct/

https://wordsandbuttons.online/interactive_mnemonics_for_dot...

In essence then, not as confusing to the beginner who might even know what a dot product 'is' operationally but not what it 'does'.

So level 1 'it's just a normalized dot product', level 2 more immediately intuitive: 'is arrow 1 pointing in the same direction as arrow 2?' or 'how close is arrow 1's direction to arrow 2's direction?'

Now what's left after that is 'Why is it so? Why did we decide on this in embeddings?'

> does no one not find programming fun anymore?

Author needs to get into Bret Victor. Has no idea how much more fun he could be having.

Programming is a step on the way to access to the state space of information. When we get to that stage, programming will seem like a maze of syntax, that has its own idiosyncrasies that force you into corners or regions in the state space, just like any DAW plugin or 3D tool, or any tool at all that exists.

They're not fun enough for a name like that

Someone needs to Tufte-fy this. I don't know if it's the hug of death or not but when I zoom out, I don't see (what I assumed would be) a distribution of deaths, a sort of heatmap of occurences. Then if you add a time dimension with color or some other parameter, it could give us a sense of what areas are most dangerous to avoid. At the moment the pins / tooltips are not very ergonomic and they obscure more than they reveal.

Useful to know, thank you!

Exactly... I've found GPT-4o to be good at OCR for instance... doesn't seem "blind" to me.

Assuming signups = $, true. But not if there's some free trial or something and then you've got a conversion rate on top of that.

Kinda makes you wonder... why don't they just advertise with those odds?

which feynman lecture is he referencing (on diffusion)?

edit: there's one called "#43 Diffusion (5/1/62)" here: https://www.feynmanlectures.caltech.edu/flptapes.html#restor... so I assume that

Excellent point, but slop takes brain rot to new heights of mindlessness.

+ MacOS + iPadOS

Well in this case, Apple is going to avalanche the slop to an unprecedented degree.

Hopefully someone reproduces results with code... cant find any code they shared

I couldn't agree more. We need LLMs that don't sound like anodyne predictable woke customer service agents.

I always make this argument:

If a human read all the text GPT read, and you had a conversation with them it would be the most profound conversation you've ever had in your life.

Ecelcticism beyond belief, surprising connections, moving moments would occur nonstop.

We need language models like that.

Instead, our language models are trying to predict an individual instance of a conversation, with the lowest common denominator customer service agent, every time they run (who to his credit, can look things up very well).

And I don't think fine tuning this "tone" in would be the way to go. A better way would be to re-Frankenstein the existing ones architectures or training algorithms to be able to synthesize in this way. No more just predicting the next token.

Well functionally it might suck a bit, but theoretically the idea of a laser is interesting, the design team in general is pretty good, I'm talking about aesthetics only here. Their device is beautiful. The team has an Apple flair certainly. I know I've seen the demos and it is surprising how slow and wrong it can be, etc. etc. But it's v1. The only good counter is they had a shit load of VC money, that's true. But still, I just don't think people should be quick to throw it in the bin. There's a baby in the bathwater, and one can't help but think if it only had a chance what could it have become.

Don't think they should be selling... just talk to users, and keep iterating on it. It'll be good, it has potential, the design is good.

For anyone who has read the paper, have they provided code examples or enough detail to recreate this with, say, Llama 3?

While they're concerned with safety, I'm much more interested in this as a tool for controllability. Maybe we can finally get rid of the woke customer service tone, and get AI to be more eclectic and informative, and less watered down in its responses.

That seems to go back in the direction of needle in the haystack again

Someone needs to come up with a "synthesis from haystack" test that tests not just retrieval but depth of understanding, connections, abstractions across diverse information.

When a person reads a book, they have an "overall intuition" about it. We need some way to quantify this. Needle in haystack tests feel like a simple test that doesn't go far enough.

Thanks for the feedback! We reduced the pricing to $20/month. I agree those features would be cooler, will definitely explore

Thanks for the feedback! Brightened the text for now on our site + landing page, but it's probably still too dark. We plan on supporting light mode soon and giving people that toggle, will keep improving in general too.

Still very useful to know, thank you for sharing. Will make the text brighter in general.

Good point, we could disable the fade on mobile.

The location of the gradient is parametric based on where the found keyword is. So if it's at the bottom, the gradient will be brighter there by default, or if the keyword is at the top, it'll be brighter there.

If there's more than one match in a single preview, the gradient accommodates for that as well. So if your keyword is "potato" and it shows up at the top and bottom, there is no gradient (or it's very small such that there is none)

There's typically enough text to discriminate (even before hover)

When the user hovers, the area that isn't bright gets brighter.

Do you mean the search text preview? That gradient is intentional and when you hover over it (inside the actual site), it brightens up.

Does this only work on Chrome? Can't seem to make it work on Arc.

location scouting for filmmakers

A better name would've probably been FastLoRA or something

I agree but "worst ever" still feels off for this reason