mrbungie

6 ms·

91.
by mrbungie 10mo ago
It is stress in the consumer economy, everything is flocking and concentrating around AI related hardware due to better ROI.
92.
by mrbungie 11mo ago
Welcome to the era of complex relationships with the truth. People comparing MCP to k8s is only the beginning.
93.
by mrbungie 11mo ago
Well, depends on the software project itself and where you are in its development lifecycle but: - (1) A lot of developing can be just chores around managing scaffolds and repeatable work, and due to this macros, autogenerated code and othe
94.
by mrbungie 11mo ago
> go through much more than just predict-next-token training (RLHF, presumably now reasoning training, who knows what else). Yep, but... > To say they LLMs are 'predictive text models trained to match patterns in their data, stat
95.
by mrbungie 11mo ago
Sad as it defeats the purpose and spirit of libgen. Being more concrete: I don't think they made libgen hoping the content would be used to create an even worse walled garden for knowledge.
96.
by mrbungie 11mo ago
It's fun when then you read last Nvidia tweet [1] suggesting that still their tech is better, based on pure vibes as anything in the (Gen)AI-era. [1] https://x.com/nvidianewsroom/status/1993364210948936055
97.
by mrbungie 11mo ago
Just as expected from a big firm with slower organizational speed. They can afford to make those mistakes.
98.
by mrbungie 11mo ago
The most fun fact about all the developments post-ChatGPT is that people apparently forgot that Google was doing actual AI before AI meant (only) ML and GenAI/LLMs, and they were top players at it. Arguably main OpenAI raison d'êt
99.
by mrbungie 11mo ago
Fair point. I don't know how to actually respond to this one without an objective measure or at least proxy of a measure on the sentiment of the discourse and it's public perception. Anecdotically I would say we're just in a
100.
by mrbungie 11mo ago
> It surprises me how hyper focused people are on AI risk when we’ve grown numb to the millions of preventable deaths that happen every year. Companies are bombarding us with AI in every piece of media they can, obviously with a bias on
101.
by mrbungie 11mo ago
I know what you mean, but LLMs are just a tool. Probably the joy is actually taken out by some form of pressure to use them even when it doesn't make sense, like commercial/leadership pressure.
102.
by mrbungie 11mo ago
It is, but biased evidence, as he's both directing and checking that frontier LLM output and not everyone is Terrence Tao.
103.
by mrbungie 11mo ago
Obviously, not the best plot to use according to Data Visualization theory and common practice, but I think it candidly conveys the point anyway. As someone else points, the data is the worrying aspect, as it points towards state-of-the-ar
104.
by mrbungie 11mo ago
You don't, but oh boy, the experience is worth it. Bazzite[1] has it quirks but it mostly works fine in desktops. [1] https://bazzite.gg/
105.
by mrbungie 11mo ago
Mirror: https://archive.is/eh7Vs
106.
by mrbungie 11mo ago
Yep, tech is incredibly circular. Once Nvidia gets there is highly probable that "disruptive" competition will apear due to mere desire/pressure for more freedom and options (and knowing NVDA, also costs).
107.
by mrbungie 11mo ago
> They used a c5.4xlarge that has peak 10Gbps bandwidth, which at a constant 100% saturation would take in the ballpark of 9 minutes to load those 650GB from S3, making those 9 minutes your best case scenario for pulling the data (withou
108.
by mrbungie 11mo ago
OpenAI would/could say the data is biased (maybe even purposefully).
109.
by mrbungie 1y ago
> Worst case, when the bubble pops, we get a glut of power and compute which (ideally) should reduce overall electricity and the obscenely bloated cloud bills. However, at the rates that demand is growing, this seems unlikely. You are ma
110.
by mrbungie 1y ago
Yeah sure, as long as you have a lot of resources for testing widely. Still, if you were to make an analogy you should target for a few devices that represent the "average", just as its done for (most) pop music production.
111.
by mrbungie 1y ago
I've heard that one too!, but seriously, it could keep going for a while.
112.
by mrbungie 1y ago
Also RIP any take-home assignment that depends at least partially on writing prose/essays. Schools will need to reinvent themselves in some ways.
113.
by mrbungie 1y ago
Nothing wrong with that, but at (1) least reference it or (2) define it yourself explicitly.
114.
by mrbungie 1y ago
Yep, and their only reference to the word points to a survey that does not mention slop even once (A survey onllm-generated text detection: Necessity, methods, and future directions. Computational Linguistics, 51(1):275–338, 2025., https:&
115.
by mrbungie 1y ago
I don't know if not getting the idea right, but I'm pretty sure people refer to AI outputs as "slop" not due to (only) repetitiveness. According to some sources: [1] Wikipedia > AI slop is digital content made with ge
116.
by mrbungie 1y ago
> It just goes to show the difference between best practices in cloud computing, and what everyone ends up doing in reality, Well, inter-region DR/HA is a expensive thing to ensure (whether on salaries, infra or both), specially whe
117.
by mrbungie 1y ago
> Sure, but isn't that moving the goalposts? It can be considered as that, sure, but anytime I see Lecun talking about this, he does recognize that you can patch your way around LLMs, the point is that you are going to hit limits ev
118.
by mrbungie 1y ago
Pretty sure you can fill a room with serious researchers that at the very least will doubt about 2) being solved with LLMs, especially when talking about formal planning with pure LLMs and without a planning framwork. PS: So just we're
119.
by mrbungie 1y ago
Yep, just a delayed combination of the full cyberpunk genre.
120.
by mrbungie 1y ago
That's not any employee, its their VP of Science.

More ›