lucfranken

It is really not that far out.

"Those iot devices are not powerful enough" is not the whole story.

Yes, big models will not run on slow gpu's and processors this year.

There are 2 things in the article: 1. Drop in a model by itself 2. Drop in a model with tools (see part: "As a sanity check")

1. Assumes that the whole "system" deployed would be stupid at the start. The article mostly describes a random "drop model on one device" and let's see how far it gets.

2. In real scenarios, the small models have cheat sheets for solving their tasks. That gives AI and the speed and performance of pre-made scripts, tools and upfront knowledge.

Add: Improve the prompt for the next device by your learnings. And there you have some flexible but also constantly improving system.

Is it slow: Yes. Does it run into many dead paths: Yes.

But a network of those devices has time, they can run 24/7.

And yes: True, Anthropic could in a financial position where there is value to post this.

And yes: It would be a wonder if scripts + flexible ai wrapper are not yet being tested in the real world.

Ah wow, they got down quite a bit already! Here it's around € 20k it seems. Not bad.

> The Netherlands of all people should understand about sea level rise and the real world capacity available to absorb CO2.

That's interesting!

It will be a huge issue which we have to face, organize and regulate.

For sure, we humans cannot handle unlimited amounts of messages, questions and other things.

But it also results in new ideas and experiences which shape humans. And it inspires humans to consider new adventures which is the good side.

Maybe we are already feel it like spam while others have that feeling of a shiny new experience in their area of expertise.

Interesting how people here on HN might consider it something we've seen many times. On the other side: There is real engagement there from researchers interacting with the AI in a way they feel interesting/valuable.

We should not forget the way broader levels of knowledge on topics around the world compared to our own small world.

It's really hard to see anymore where the real money is made and what is a big facade. As it's really simple now to create a huge facade. Before it needed more work to show something big which was actually small. But now you can generate the show.

And there are actually some which are really big. But there are so many more overflowing the world with non existing businesses that it's hard to decipher what is real and what is not. It also makes learning from the successful ones way harder.

Isn't that how they just release all features? So they can do progressive roll outs and telemetry on issues with it?

Not sure if they later move the code from inside the flag check to the main code or that they keep the flag check.

But if they would keep all features behind a flag that would not make most sense as you then would have not many features without telemetry.

Real world stuff, physical stuff, real services are still complicated.

Moving around papers, data and that kind of jobs becomes automated.

The real math problems etc: I think it will create many more interesting problems and challenges. It's not like something is solved and nothing is left. That's the wrong way to look at it I think.

Not sure on that, maybe the options to choose from will be generated and curated. Same as we do with tagging datasets for images. Might be wildly successful for real world decisions.

Blackberry wants it's keyboard back :)

Interesting that there is no mention of running your Bots and Agents directly from this device. It's interesting to think about whether we will keep managing so many apps on a device or that all of those will be merged for the user interface.

Jev is such a different approach where you have to be specific about what you want and which options are open. Really interesting how those things evolve in usable features for people.

Also with this example the speed of new launches based on a launch is just incredible.