yeck

7 ms·

31.
by yeck 3y ago
By that do you mean that you believe AI is safe by default? Or that AI safety is not possible?
32.
by yeck 3y ago
I've seen some likening Sam to Caesar, being betrayed and such. Certainly there are communities that match what the parent comment describes. HN has generally been more balanced than I would have though, though. The fan boys I've
33.
by yeck 3y ago
Not OP, but the choice of phrasing in a number of reports implied that the board was uncertain of their decision. Normal sensationalization if the same limited information that was publicly available at the time.
34.
by yeck 3y ago
Anyone find any sources with new information? Now that we are well past the 5PM (PST) deadline I'm curious if anything has emerged.
35.
by yeck 3y ago
That's 4x the amount of buy-in than is required to fire most people.
36.
by yeck 3y ago
Aspects of this article were surprisingly good. This part was not one of them, though. What's already been released publicly is feels "rooted in the realm of science fiction".
37.
by yeck 3y ago
You are chaining some big assumptions together to make this conclusion. We can suppose AGI is around the corner, but these other assumptions are massive leaps that would need strong arguments to back them: - AGI thinks similar to humans - A
38.
by yeck 3y ago
If Microsoft ends up essentially at the helm of OpenAI after this all this then I'll know we are living in a tragic comedy.
39.
by yeck 3y ago
Yeah, I suspect there will be a strong over reaction regardless of whether Altman is reinstated or not.
40.
by yeck 3y ago
To be fair it is a very subjective term, god-like. You could make a claim for many different technical advancements to represent god-like capabilities. I'd claim that many examples exist to day, but many of them are not readily availab
41.
by yeck 3y ago
Hah, yes. I chuckle to myself every time.
42.
by yeck 3y ago
I suspect some do and some don't. Hard to know what the ratio is.
43.
by yeck 3y ago
I feel that people have a right to life and liberty, but liberty does not mean access to god-like powers. There are many people that would do great things with god-like powers, but more than enough that would be terrible.
44.
by yeck 3y ago
I wouldn't call GPT-4 sub-human intelligence. While it's intelligence is less robust aggregate human intelligence, I don't think there is any one person alive who can compete with the breadth of GPT-4 knowledge. I also think
45.
by yeck 3y ago
> models to be (easily) swayed is a different problem No, this is the alignment problem at a high level. You want a model to do X but sometimes it does Y. Mechanistic interpretability, one area of study in AI alignment, is concerned with
46.
by yeck 3y ago
Was the article changed for this? Used to be this one from the verge: https://www.theverge.com/2023/11/18/23967199/breaking-openai... but was since changed to https://www.forbes.com/sites
47.
by yeck 3y ago
Isn't that a massive quality improvement though? How many applications for LLMs are not feasible right now because of the ability for models to be swayed of course by a gentle breeze? If AI is a ship with a sail, data is the wind and a
48.
by yeck 3y ago
Until I can trust that when I send an AI agent off to do something that I will be successful without me babysitting and watching over it constantly AI won't truly be transformative (since the human bottleneck will remain). This is one
49.
by yeck 3y ago
> the runner is closed source My team runs our GHA runners on our own infrastructure using Actions Runner Controller (ARC). > The fact that you have to push a commit to the repo to test a change, and then wait for a runner delays ever
50.
by yeck 3y ago
I've seen this sentiment mentioned many times. That at school "abuse is more likely to be noticed". But how much more likely? How likely is noticed abuse to be resolved in a meaningful way? As an anecdote I have close family
51.
by yeck 3y ago
As users' codebases grown over time, the average usage of users will probably continue to grow. By a static price increase alone won't be a great model. They need to make their models and infrastructure more efficient first. Or cr
52.
by yeck 3y ago
Controlling for academic experience probably raises the average accuracy of labelling, but by how much? Clearly having a degree will not make you omniscient in your major, let alone other subjects.
53.
by yeck 3y ago
K8S maintains the internal state of the system in JSON which is easy to convert to YAML which is seen as more readable in comparison. I think their choice to support JSON and YAML as an interface for configuring K8S is because of this. I&#x
54.
by yeck 3y ago
Code also requires a runtime or needs to be compiled to something static, like YAML or JSON, in order to be consumed. That latter option is essentially how things are today given that templating tools the generate configurations are plentif
55.
by yeck 3y ago
Bun is getting a lot of hype due to all the press, but almost everyone is just echoing Bun's docs and marketing material and lack direct experience using it. I haven't done a deep dive into Bun yet, but I have tried porting a lega
56.
by yeck 3y ago
The gist of this is that ff those who are concerned with AGI catastrophe are wrong but try to prevent the perceived looming disaster then they might do bad things. Bad because they caused harm to people by were unnecessary. I don't thi
57.
by yeck 3y ago
The funny thing is that a lot of the ppl on social media that I've seen taking this stance against "longtermism" tend to also hold their own long term agendas about utopia. This "conflict" feels really contrived to
58.
by yeck 3y ago
> health benefits are dubious Are they? Certainly if you stand still for hours I'd be skeptical for that, but while you are standing at your desk you are able to do all kinds of movements that you are not while sitting. And you are
59.
by yeck 3y ago
Upgrading to a WI-FI mesh that worked well for my home has had the largest impact for me. I've always liked to be flexible with where I'm working in the house but many spots had very degraded WI-FI signals. Working in the backyard
60.
by yeck 3y ago
Yeah, it is harder than other things, but if we can train a model to explain collections of pixels in human language then we might be able to do similar with collections of activations. I don't know if that is the direction, but just a

More ›