jazzpush2

8 ms·

121.
by jazzpush2 7mo ago
It's everywhere on YouTube, usually as a 'hidden' ad in the alt-right manosphere (e.g. the recent Nick Shirley video he wears a sweatshirt for a gambling site throughout, with constant name drops of it that aren't over a
122.
by jazzpush2 7mo ago
Right. If these are the people making the decisions (and don't even get me started on the 'technical' folks at a16z...), the cluely-esque enshittification of VC over the last few years makes A LOT of sense.
123.
by jazzpush2 7mo ago
That's an absolute insane amount of code 'created', but the natural follow up is: for what? Are there examples of what this software has created?
124.
by jazzpush2 7mo ago
And allow more insider trading.
125.
by jazzpush2 7mo ago
Who is NVIDIA killing, exactly? Is Apple complicit in killings because operators planned missions on Macbooks? Dell? Microsoft?
126.
by jazzpush2 7mo ago
Treemaps are used to show hierarchical data. But here he doesn't even bother to show more than one-level of hierarchy. If I want to find, e.g. Police, it's near impossible, since I have to scan with my eyes (when, again, it's
127.
by jazzpush2 7mo ago
You think, charitably, that this is an astroturf, really? What's the distribution of upvotes look like for front-page posts, binned in 20 minute intervals? Reviewing your post history, it's overwhelmingly in non-tech related thr
128.
by jazzpush2 7mo ago
This is a surprisingly bad treemap, which is surprising given how easy it is to build a good one these days.
129.
by jazzpush2 7mo ago
Amazing. A very cool niche area, dataviz x ai/ml. See also: - mlu-explain.github.io - visxai.io - google PAIR's explorables - GA Tech's poloclub.
130.
by jazzpush2 7mo ago
Do you think the investors of xAI want this behavior baked into the model? Do you think other frontier labs enforce their models to praise their CEO and never insult them? And, how does this fit into a vision, exactly? What vision might tha
131.
by jazzpush2 7mo ago
What? I work for a different frontier lab now, and it absolutely would be ridiculous if they told me the same thing. Luckily they haven't. What products have you worked on where this would be deemed normal?
132.
by jazzpush2 7mo ago
I.e. time American tax dollars to save the day!
133.
by jazzpush2 7mo ago
Karpathy makes great educational content. It's not clear what industry (or academic) research he did even now, five years later.
134.
by jazzpush2 7mo ago
In an interview with xAI I was literally told that certain parts of the model have to align with Elon, and that Elon can call us and demand anything at anytime. No thanks!
135.
by jazzpush2 7mo ago
What I don't understand is why Congress is doing nothing. Today we were told Trump went to war because he 'felt' we were threatened by Iran. Absolute BS. See video here: https://x.com/The_Kremlinn/status&
136.
by jazzpush2 7mo ago
You'd think they could just use AI to make Jira not take 5 seconds to create a ticket. Linear does it instantly. Just use AI to create a SPA version.
137.
by jazzpush2 7mo ago
I have several friends at the big 4 who work ~2 hour weeks.
138.
by jazzpush2 7mo ago
Kennedy, Bush, Clinton, Newsom related to Pelosi, etc... This guy might be onto something!
139.
by jazzpush2 7mo ago
This is why skateboarding is a great hobby. You learn from a young age that falling is normal and necessary for progress.
140.
by jazzpush2 7mo ago
All of the links I posted were published before ChatGPT, no AI.
141.
by jazzpush2 7mo ago
Good call, I'm sure I've left off a lot of amazing work, and his is certainly top class!
142.
by jazzpush2 7mo ago
I already have an Anthropic sub. There is not a single Democrat candidate I agree with or would gladly donate to. That party is almost as much of a joke as the Republicans are, and they all fuck kids anyways. Besides, that money would only
143.
by jazzpush2 7mo ago
Just cancelled my chatgpt subscription.
144.
by jazzpush2 7mo ago
I like the idea here, but the final product is just so far from what good interactive articles/explanations actually look like. E.g., this style of article: - https://mlu-explain.github.io/decision-tree/ - any art
145.
by jazzpush2 7mo ago
It's very bad in my experience. It hallucinates like crazy - e.g. something simple as enumerating the correct hidden dimension for a transformer-based model (same across all layers) it gets wrong often.
146.
by jazzpush2 7mo ago
Why? Genuinely, who cares? Is some demographic group not caught in the 8 going to be offended by basic checkbox screen? Is someone with a niche form of colorblindness going to have difficulty navigating the UI?
147.
by jazzpush2 7mo ago
That's nothing. Former/current YC founders are also abusing BookFace. I did YC and now work at a frontier lab. I've received multiple spam-style emails from (mostly young) current founders tagging me and all other YC-alum at
148.
by jazzpush2 8mo ago
I don't get it. The author is 'honest to a fault,' but then lies about discussing the polygraph with others and posts an anonymous op-ed.
149.
by jazzpush2 8mo ago
A16z partners don't know shit. Brain dead nepo-babies - how's Cluely?
150.
by jazzpush2 8mo ago
This post title is completely misleading. From the article: the shooter's behavior triggered internal alarms, and some employees asked leadership to alert authorities, but: | OpenAI leaders ultimately decided not to contact authorities

More ›