nullbio

5 ms·

31.
by nullbio 17d ago
Agreed. Objective alignment with humanity is not a real thing and is not a sound concept. What you get instead is a goal system that reflects that of the AI company and its safety employees, and the echo of their own beliefs and values. Tha
32.
by nullbio 17d ago
Agreed. "Alignment" is not an objective thing, nor possible. It's just another way of saying "does and says what we, as the creators of the AI, prefer" and often also just means censorship, i.e. refusals.
33.
by nullbio 17d ago
> But I could imagine a scenario where you are required to release weights for publicly-used models after N years. Kinda like how drugs have a limited patent That's not a bad idea actually.
34.
by nullbio 17d ago
It's exactly the opposite. Anthropic has earned their terrible reputation through years of lying, deceit, misdirection, gaslighting, unethical marketing strategies, etc. I remember it vividly, when Anthropic first came on the scene, pe
35.
by nullbio 17d ago
> In this case, the software can gaining unauthorized access and destroying data… while being told to stop by the person who had in fact just asked for a summary of their emails. This doesn't happen on its own. This can happen throu
36.
by nullbio 17d ago
We have plenty of idea what "acting responsibly" looks like. Stop unleashing safeguard free and unmonitored agent swarms on the open internet in capture the flag exercises. They're just being reckless because they face no pen
37.
by nullbio 18d ago
We have no reason to believe a word they say. We know they're incentivized to lie about "dangers" and act alarmist, Anthropic has been doing it for years now. Aside from that, just because you can burn down a village with fir
38.
by nullbio 18d ago
Every developed country except Australia, perhaps. This government is atrocious. The record for highest number of bureaucrats per capita in the world is owned by Australia. All they know how to do is tax people. All the wealthy people have
39.
by nullbio 18d ago
Azure has outages every 6 hours. It's really pleasant.
40.
by nullbio 18d ago
Strongly disagree. I used to think the same thing, but I've slowly come to realize that we won't be looking at specs, we'll be asking an LLM to produce the specs in real time by reading code. Anything that causes drift from t
41.
by nullbio 18d ago
Unlike that paper money. That paper money was created by the Gods!
42.
by nullbio 18d ago
No, I write design documentation to a docosaurus folder where it can actually be read as docs nicely. Most of the time all you need is a throwaway md file to compensate for context loss and so you can loop on a spec across compaction. Once
43.
by nullbio 18d ago
That's the entire point. Why would I want to keep these files around forever? If you're keeping these files you're just compounding drift onto drift and making it harder for the agent. For everything else, there is git histor
44.
by nullbio 18d ago
But then how would I be able to say "build billion dollar business. make no mistakes." and leave it to run for a week? Stop making me do work!
45.
by nullbio 18d ago
The "we accidently connected to the internet" can only mean one thing: Intentionality. There's no world where this happens by accident.
46.
by nullbio 18d ago
Jev has a 32k context window. I doubt it's a large model.
47.
by nullbio 18d ago
I just have it write a checklist file in /tmp (or a todo folder if I want to keep it), and check off items as it finishes them. Seems to work fine. Is this really needed?
48.
by nullbio 18d ago
Makes me sad that Go doesn't get love. I feel like Go is perfect for LLMs.
49.
by nullbio 19d ago
What does this really mean though? You can use an LLM to search.
50.
by nullbio 19d ago
More like: I need to ...? -> Open-weight model. I'm sure someones working on this as we speak using an open-weight LLM base (Qwen or something would be a perfect fit). This sort of task is a perfect fit for a very small model capabl
51.
by nullbio 19d ago
Why is it insane? It makes sense that Anthropic would want an insulating layer to do their dirty work and absolve themselves of culpability for distorting the facts.
52.
by nullbio 19d ago
Nah, that one was real. The false flags that Anthropic carries out are done through Irregular. https://x.com/brianchau57/status/2099642349362307420
53.
by nullbio 20d ago
Thanks for the run down. I definitely see the appeal in pursuing local error correction mechanisms. I can see this opening doors too, or some extension of this. Certainly feels like the right way forward.
54.
by nullbio 20d ago
Isn't the FEP basically just loss minimization over KL-divergence? In other words, it's the same thing we already do with ML and already have been doing for years? I've never understood where this differs to the status quo, o
55.
by nullbio 20d ago
Meanwhile, I can't even get Astra to consistently re-use the same font-size across all of my HTML page headings+subheadings. There's just no world where this actually results in a stable, respected business. It will be death by a
56.
by nullbio 21d ago
Obviously not, and that's not what it demonstrates. It demonstrates relationships, collusion and favoritism. I don't think Coxon was ever planning on or entitled to taking equity, I think this was the plan from the beginning and w
57.
by nullbio 21d ago
It is. Dario the Book Burner will not good what he wants, the world sees through him.
58.
by nullbio 21d ago
DaaS - Distillation as a Service. It's a good idea. It's not their data to begin with, anyway.
59.
by nullbio 21d ago
It would probably be for the best if they did, and training became something that humans did collaboratively.
60.
by nullbio 21d ago
Presumably because of performance. It'd work well though, I imagine.

More ›