i5heu

Looks like 27b models are in reach.

“In the current generation, our density is 8 billion parameters on the hard wired part of the chip., plus the SRAM to allow us to do KV caches, adaptations like fine tuning, and etc. In our next generation, we would have the ability to go up to 20 billion parameters in a chip. Even with trillions of parameters, we’re talking about few tens of chips, which is a very, very small compared to anything else out there on the market today.”

https://www.nextplatform.com/compute/2026/02/19/taalas-etche...

Edit: i do not know how reliable this page is... it has a lot of typos.. for me it does not look like it was written by a LLM

Looking at their career page it looks like they do not care that much about PR at the moment... considering that they have a live chat with 14000t/s via Llama 3.1 8B[0] i don't really think they need to do PR either.

So i guess maybe they currently try to solve a very hard problem with a small focused group before scaling or they are dysfunctional.

Also Llama 3.1 8B is a dense model AFAIK and they are fast by nature. As there are not a lot of dense models these days i could imagine that they try to optimise for MOE models.

[0] https://chatjimmy.ai/

This Kimi K3 had only one prompt, because I was using the free tier and it ran out of free credits. So I was not able to go on a Journey with Kimi K3. I did with Fable, which is why some elements on Fable look good… for instance there is a partially complete pyramid in Fable that I added with additional prompting.

This should be the first thing on this page.

Also I imagine it pretty difficult to get good data from that because of all the muscles that do stuff if put in water and you would hold your breath.

There is no way people will put up with that.

I think it is because they want something that feels nice so people will go there more often?

So they get more data of the same person over time.

And slide 25, although it is questionable if this is really the result of such a device like Midjourney is presenting.

If it is then wow!

At least in Germany we have “Bio” which is a organic label that is controlled at least somewhat.

It was just an example. You probably can imagine some kind of level you would consider definitly conscious.

I recomend focusing on designing Tests before rigid definitions of states of tests. - it is much clearer that way what is asked from something to be conscious.

But again these are all just indirect tests because we cannot test the core of consciousness because we don't know the core of consciousness.

A lot of this is because food is hard.

For example: people in developing countries throw away more food then in developed countries although it is relative to their income much more value. The reason is because they often do not own fridges, use a lot of rice which spoils fast or have difficulties with the food supply chain in other forms.

Food waste is a bad indicator for food value.

I’ve seen definitions that I like they where in the direction of „if a system can recognize a problem it has not encountered before and can attempt to solve it with onto the problem adapted solutions, then it is conscious“

But then again this is just a external crude form of test that can lead to something like „light bulbs emit warmth, so fire must be a light bulb“

It might not even need a transformer.

A specific pattern of self-referencing data could be seen (or not) as low-level consciousness in the future, when we know what consciousness exactly is.

It might be that stockfish is already something future scientists would define as "conscious".

Altough it is diffucult for me to Imagine that specific example.

So, if I understand this correctly, this paper proves that LLMs can run on crude VMs?

This is not at all how all of this works.

If you train an AI in one thing it will become better in the other.

What information goes out of the system which is novel?

I think there is very little truly novel information. Most information including the information of "breakthrough idea after mis-hearing someone" is just a mix of previous information.

I guess you are already aware of that... just for completness sake.

And humans just work with some electrons that are fired via neurons.

If you limit yourself to reductionism you will not see emergence and thus human consciousness is impossible for humans.

I would like to read it!

Please publish it!

Yes I did and it is no problem at all.

You have to consider that you do not own the product from bambu.

So to say you spend significantly less money for a product that will be changed whenever bambu sees fit to whatever extend they see fit.

I regret my bambu purchase a lot. I have to keep up with all the ways bambu wants to lock in my hardware and take it basically away from me.

AFAIK there is some inflammation potential if earplugs are used all night everyday.

https://www.nytimes.com/wirecutter/reviews/advice-wearing-ea...

You just use docker.

It is like 4 lines of config for Postgres, the only line you need to change is on which path Postgres should store the data.

Hi,

so i am not entirely sure what you mean with "trying to be an intellectual", to me it sounds like a personal attack tbh and a quite cynic one.

I wanted to preventively write about all the concepts i have looked at because i wanted to prevent that people get hung up on the form and stop discussing about the content the the core message of my post.

As i wrote i don't see that "150 years of Peirce" have produced a model that is similar to what i am trying to say here.

You apparently take a offence in whatever you are offended and i want to give you a chance to speak with me directly about whatever you seem to think i have missed under the precondition that you try to actually engage with my text and don't hand wave it in a cynical fashion like you did until now.

I have looked, as i wrote, at Peirce and although he describes theories that are similar to the end effects of the DDS with AI-skew, it still is something completely different.

Besides that i am saddened that you seem to put form over content and that you did not really engage with the blog post itself.

Thank you very much for this comment! I was not aware of that.

Just watched the end scenes and this more like an execution in a heroic light.

I have not seen something like this in this significants, vibe and presumed statement in western media. There are a many examples of uncelebrated murderes of “the bad guy” which the hero does not want or does “against their will” but I have seen no western examples of this kind of celebration of killing the bad guy.

I think that >in the future< this will be a non problem as there is reality itself that is a much better validator for behavior then human text.

We already see this with synthetic training data that basically uses logic in form of math and code as constraint.

I love my sauerkraut! Self made is also so so much better then industrial sauerkraut

One album is the price of an entire month of streaming.

The road to death with any OSS project is scope creep.

I appreciate your comment though! Thanks :)

I will update it later down the line as YT changes its layout and for bug fixes... but i think this was not what your question.

I don't think that something super bad can happen with these uBlock filter, they will sanitize the filter heavily.

Maybe a potential attack vector for these lists in general is to hide the body of a few sites but this is more annoying then dangerous AFAIK.

Mhhh although this is a bit OT it is also very interesting:

Germany has unions and works council. It is required by law that companies allow works council to exist and if they exist they get certain rights.

In Germany, they serve two functions. The first is called co-determination, through which works councils elect members of the board of directors of German companies. The second is called participation, and means that works councils must be consulted about specific issues and have the right to make proposals to management. ( https://en.wikipedia.org/wiki/Works_council#Germany )

Also Germany has good amounts of regulations for certain issues (sometimes too much regulation but oh well).

And Germany has many options to participate in local and regional politics that prevent the worst offenders.

But in the end a company is still a company that is forced to act in a way that maximizes it's profits. I think from my perspective we do not have such wild predator companies like you see them in the US but there are certainly a few very dubious things going on.

Except maybe the gGmbH and Vereine (clubs) which are companies and semi company structures that must act for the common good and without profit interest.

And one funny thing is that in Germany stock companies are required to act to the "best interests of the company" and not the "best interests of the stock holder" - in German law the company includes the worker, the future of the company and social aspects.

I think it is because of the Chinese new year. The Chinese labs like to publish their models arround the Chinese new year, and the US labs do not want to let a DeepSeek R1 (20 January 2025) impact event happen again, so i guess they publish models that are more capable then what they imagine Chinese labs are yet capable of producing.