Well, considering that the long term idea is to have AGI, general intelligence, it seems that the goal as also to only have a single product in the end.
There may be different ways to access it, but the product is always the same.
Not easy to make the connection to Palmer being fired here. The case was decided much later and it equally involve John Carmack, who stayed at Meta much longer, until he left out of his own accord.
The logic is still the same. If the VAE is trained so that it is biased toward human preference, then the probability of false positives in real world images would increase.
Generally, the VAE is mapping from a small latent space to a large image space. This means that there must be a large number of images for which no reverse mapping exists.
It should be possible to identify images that have not been generate by the VAE since they are not part of the set images that the VAE can generate. The other way round is a bit more difficult as there may be images that can be mapped to the latent space and back without loss but have been generated in another way
From what I gather, this project started out as an implementation of a code-interpreter using a local LLM. Basically your instructions are used to write code by the LLM, which is then executed. The idea is that it can be much more powerful having access to your native systems shell instead of only sandboxed python.
In the meantime, it seems that also models with vision capability have been added, that can be used to access GUI based applications, not only the shell.
It's a very exciting concept that lives in a space where open source software should have a significant advantage due to its transparency. (Or would you give a black box device access to everything on your computer?).
It also seems to one of several emerging projects that try to sketch out a path for ideas of how LLMs could change the way we interact with computers.
Is it uncommon to dislike the latent smell associated with cooking? I love cooking, but I go as far as changing my clothes after it to get rid of the smell. I could never imagine having an open kitchen.
It's as if someone created one element that is perfectly suited to build microelectronics. Sure, there are other materials that improve on one property or the other, but there is not a single other element which balances properties as well as silicon.
Not even mentioned yet:
- Excellent mechanical properties of the single crystal (think MEMS, or wafers that don't break all the time)
- Piezoresistive properties can be used to measure strain (also quite unique due to silicon band structure)
- Optical properties perfectly suited to detect visible light (think detectors, image sensors). Good combination of band gap and carrier lifetime to build solar cells.
Silicon oxide grown on Si is actually amorphous, so it is not lattice matched.
But you are complety right, the oxidation properties of Si are really fortunate and ICs would have taken decades longer if it were not for that. SiO2 is really the unsung hero of the silicon age.
- SiO2 has a high bandgap and a very good insulator.
- It is quite inert to many chemical and gasses. (e.g. germanium oxide is soluble in water, which is a headache)
- It can easily be grown on stoiciometric form by oxidizing silicon and will form an abrupt interface to Si.
- The formation proceeds by diffusion of oxygen to the Si interface. This is in contrast to other metal oxides, where the metal will diffuse to the surface and create a nonstoiciometric mixture.
There is no other semiconductor that forms as good an oxide. Very few metals form insulating oxides on their surface, one notable exception is Aluminum.
Edit: The famous paper that describes the SiO2 formation kinetics was actually co-authored by Andy Grove, from intel CEO fame.
Pretty wild how well GPT4 is still doing in comparison. It's significantly better than their model at creating compilable code, but is less accurate at recreating functional code. Still quite impressive.
Call me ignorant, but I am extremely put off by products that blatantly rip off naming schemes and essentially position themselves as a copy of another product. Granted, "Orange" went through a number of itereation and is now less similar to the product it originally copied.
From the article: "Like many, it is let down by its software support..."
Well yes, no news. This was always the strength of the original Rapsberry Pi and is the reason why most of the impulse-bought copied products end up in the parts bin.
A bit surprised to see this on hn at this time.