I do mean it about it having a democratic function by the way. I think it is important that there is widespread literacy about this stuff, its limitations and power. Its necessary for good democratic deliberation on this, which we are already being asked in the UK to do.
Its interesting too that Meta's Llama is the model you can probably most readily run easily on the edge, but they haven't done this. Llama is the most open US foundation model, I think. Anyway!
It then pretended to reset its window and then admitted it hadn't. Its not clear at all how I'd do this.
Feels rushed, not extremely production ready, like so many AI deployments.
Finally in messing about with it, it is entirely unclear how to empty the memory (which OpenAI provide). I got it into a real state, asked it to delete the memory, which it did, then it recalled our prior conversation, finally admitting that it wasn't really deleting our chats.
A primary level, it is a good thing if more people have first hand experience, to form a judgement. On another fundamental level, this is laser guided to confuse an LLM with a person. You are messaging an LLM where you are also messaging your mates. This is complex to say the least.
Meta rolling out Llama to every WhatsApp client seems...interesting. I am bracketing here general concerns about AI as such.
I think that all those techniques may be in the client-side mix! I also want to try and use very minimal JS, and if one does use it, to keep it low/zero dependency and lightweight.
My solution is probably to periodically inline the whole documentation for the Go programming language, just in case...
Currently considering stuff like: if this archive suddenly didn't have access to the internet, and (let's be frank) contemporary code writing depends on the internet for documentation at minimum, what's the deal gonna be?
I am gonna be working this out soon in the context of building a very image heavy archive, where the need is for it to be ultra low carbon emissions. I shall keep you "in the loop".
Absolutely!
Ability to permanently bin a "Discover" feed is a nice feature of this here Twitter clone.
Like Wittgenstein on writing the Tractatus Logico-Philosophicus and believing he had completed philosophy, exiting to teach primary school children maths.
So Charli xcx, having completed pop, exits for a career in film.