Not sure if things have changed recently, but up until at least 2019 (last time I had to do it iirc) you definitely had to do a "flagpole" (drive through the border and immediately turn back) to change your status in Canada.
It uses Bluetooth to stream audio, but everything else happens through a WiFi connection exposed by the car that the phone automatically pairs with after the Bluetooth handshake.
> assume it hasn't had much thought or care put into it
> I'm not really motivated to even try it out
That's a little uncharitable. I get where it's coming from, because I also still feel a tinge of (hypocritical) disappointment when I realize things are built with LLMs, but I truly think we need to start moving past this. It's not a giant corporation trying to scam you out of your hard earned money by passing slop off as a lovingly crafted product: it's somebody's pet project, somebody who clearly had a fun "what if" idea and was able to materialize it over time, with a TBD amount of care/iteration/refinement. I think it's something worth engaging with. Worst case scenario it sucks, like many cheapo projects do -- best case scenario it's awesome and/or it sparks a new interesting idea in somebody else.
Holy crap, I was not prepared for how fast it responded. I just wrote "Just wanted to see how fast you are! Can you write me a quick story about a tiger who lives inside a block of cheese the size of a house?"
The magic is in the fact that they essentially have an ASIC llm device. There is no other trickery. The problem they will face is that it is actually locked in silicon, so upgrading models will be difficult, and likely require new hardware each time.
At some point I imagine you’d add a software layer on top that holds more current training and can be called as needed trading off for slower responses. There’s already work out there splitting models across networks. You could have the base on silicon, some stuff in memory on the machine, and another frontier tool in the cloud.
Taalas actually already support LoRA, basically doing exactly what you say.
The other thing that I think is really interesting about all of this, is that LLMs are already perforce behind the times with their knowledge cutoff, so adding an additional ~3 months for bake into silicon isn't such a huge deal, I think, for the ~10x more efficient and faster you get.
Yeah, really a fascinating time in computing. Once these chips start to become more common I'll be curious how people find ways to use them. One can imagine a world where really simple inference is available on dirt cheap chips found in toys and other low cost consumer electronics.
There are already studies proving that a "stupid" model with a good harness + tool calling will outperform a "smart" model.
Things like this give me hope for a system that can be fully local and private, but also with the ability to be almost infinitely extendable with tools.
Is the size of the ASIC limiting factor or can they infinitely tensor parallelize? If thats possible then it would make it only a matter of economy of scale, there are good enough models already for people to invest in that kind of platform
I pasted and instantly hit enter on this prompt: "I generated a filter set using REW v5.31.3 using real-world sweep tone measurements from the room I'm listening in . How can I use it as my MacOS output equalizer so that my spotify music is adjusted for this room and speakers"
and it gave a very reasonable answer in non-perceptible time.
idk if you're still actually looking for a solution (I got nerdsniped heh), but I found a seamless way to have it locked into the speakers + room, instead of a specific source/software, is using an iLoud subwoofer that routes to any speaker setup and has room correction + EQ that lives on the sub.
Isn't it crazy that we can send a message, across the world near instantaneously and have a coherent reply, generated by a computer, sent back to your screen, in under 500ms in most circumstances.
I find myself getting caught up in the sheer speed of modern computing and networking. The fact I can play an online game with 10 other people is just insane.
I got into Rust development via LLM last year, and being able to do things budgeted in nanoseconds is a heady feeling indeed. Real time video and audio analysis? Totally doable, plenty of time budget. 16.6ms is a long time, it turns out.
For what its worth the frontier lab models can surely be a lot faster if they wanted them to be but theyre supply constrained so theyre doing stuff like multi tenancy. Since you cant self host them no one outside the labs really knows speed as a solo tenant
You can kind of get a sense by running these things at home - I'm currently running Laguna. One interesting thing is that per stream doesn't actually slow down that much with multiple concurrents, because the bottleneck remains the memory bandwidth until pretty significant request depths, and then eventually you hit the GPU's limits. It's one of the big forces that pushes for centralization in this stuff, the fixed costs to run one are huge, the marginal costs of additional tenants, relatively small.
Fully interactive games where you can talk to every NPC by text or voice and have an LLM drive the story (with your own meta prompts to guide it, if you so wish). Maybe even have them generate assets on the fly too.
I’m still trying to figure out coding agents. I can’t even begin to imagine the things it would enable. Even the most mundane ideas like LLMs-in-HiFreq-trading have huge implications.
RSI will be models better than Fable running faster than this, you won't even need a human in the loop to figure out what to do. The high level goal will be accomplished better than the human in an instant
Long time customer! I bought one after your original post on HN. Congrats on the milestone. Honestly, it's such a good product and I'm such a bad and inconsistent piano player that I completely forget it exists for months at a time. Then I remember about it, open the app, and everything is there: my attempt to figure out a song, my kids smashing the keys as hard as they can, my cat walking across the keyboard. It's such a joy.
I realized just a couple weeks ago that I've had my Brother laser printer for 10 years, and I've only changed the toner once -- so long ago that I don't remember when that was.
How many of those are "the exact same product, but better" though? Definitely the iPhone, and possibly Slack. Everything else you listed was a drastic change in the way an existing service (taxis, home media, backups) was delivered or priced, to the point where they feel like either an entirely different product (or an unsustainably priced one that wins by virtue of being cheaper at first, see Uber and Airbnb).
I'm not fundamentally disagreeing with you -- but I think there are some things where "everybody has clearly been doing this wrong the whole time" doesn't pass the smell test: ovens have been around for millennia, including through the whole industrial revolution and recent exponential modernization of technology.
> Everything else you listed was a drastic change in the way an existing service (taxis, home media, backups) was delivered or priced, to the point where they feel like either an entirely different product
Doesn't this still line up with the original point though?
The incumbents had the exact same opportunity to capitalize. Blockbuster could've leaned into streaming hard before Netflix.
I do think that's become a bit harder recently, though, so the original point is slightly less relevant (though obviously still exists, see for example OpenAI vis-a-vis DeepMind). Big companies are much faster on chasing trends, even if they end up amounting to not much. See crypto, for example. I do think part of that reason is because the current wave of companies capitalized where incumbents faltered, so they're desperate not to make the same mistake.
Google was definitely "exact same product, but better". It was a search box.
However, I think Anova would be a better direct counterpoint. They started in sous vide, replacing existing products that did the same thing but better. But now they are literally trying to replace ovens. That effort started after the Electrolux acquisition in 2017, but they do seem to be pretty successful. A friend of mine really likes their steam oven, and if I had more space in the kitchen, I would be tempted.
Yahoo. They very nearly bought Google too, but refused the initial $1m offer and then bid too low the second time. Kind of hilarious to see how that played out over time.
Yahoo had two failed attempts to acquire Google; in 1998, Larry Page and Sergey Brin approached Yahoo to sell their nascent search engine for $1 million, but Yahoo declined the offer.[37][38]
In 2002, when Terry Semel entered into negotiations to purchase Google. Google was reportedly seeking a price of $5 billion. After weeks of negotiation, Yahoo's final offer was $3 billion, a figure that Google's leadership rejected, leading them to terminate the deal.
It's probably a bot account that tried to read the article to come up with a reasonable comment, but TCRF likely serves a "you're a bot, go away" static response page when it's accessed by bots. Pretty funny.
There do seem to be annoying false positives, but this particular account really is a bit strange. Months of silence after signing up, and then this non sequitur…
