HNHacker News
TopNewBestAskShowJobs

spencerchubb

1,275 karma · joined December 3, 2021

https://twitter.com/spencer_chubb
submissionscomments
spencerchubb··on Zod: TypeScript-first schema validation with static type inference
Couldn't this be improved by showing a diff between the mismatched types?
spencerchubb··on Addition Is All You Need for Energy-Efficient Language Models
Cheaper compute is basically a prerequisite to making better models. You can get some improvements on the margins by making algorithms better with current hardware, but not an order of magnitude improvement.

When there is an order of magnitude improvement in hardware, the AI labs will figure out an algorithm to best take advantage of it.

spencerchubb··on Differential Transformer
also presumes that the LLM knows it is an LLM
spencerchubb··on Differential Transformer
it's not "just" model error

during pre-training, there is never an incentive for the model to say "I don't know" because it would be penalized. the model is incentivized to make an educated guess

large transformer models are really good at approximating their dataset. there is no data on the internet about what LLMs know. and even if there were such data, it would probably become obsolete soon

that being said, maybe a big shift in the architecture could solve this. I hope!

spencerchubb··on The Remarkable Life of Ibelin
I don't even consider tiktok as social media. It's more for entertainment like netflix. There is not much socializing that happens on tiktok
spencerchubb··on Max Schrems wins privacy case against Meta over data on sexual orientation
if you watch enough gay content to get gay ads, then this concern is redundant
spencerchubb··on Max Schrems wins privacy case against Meta over data on sexual orientation
I wouldn't care. I might even be glad if one of those ads informed me of a lifesaving cure.
spencerchubb··on Show HN: One – A new React framework unifying web, native and local-first
I don't think it's an age thing. I'm 22 and I agree that I can't be bothered with learning javascript frameworks. There are too many other things to learn in our limited lives
spencerchubb··on Max Schrems wins privacy case against Meta over data on sexual orientation
This is a weird thing to be upset about. As a gay person, I don't want to see ads that are meant for straight people. I would rather see ads relevant to me.
spencerchubb··on Were RNNs all we needed?
When you train a neural network, it is not search, it is descending through a curve.

If you were to search for billions of parameters by brute force, you literally could not do it in the lifespan of the universe.

A neural network is differentiable, meaning you can take the derivative of it. You train the parameters by taking finding gradient with respect to each parameter, and going in the opposite direction. Hence the name of the popular algorithm, gradient descent.

spencerchubb··on YC criticized for backing AI startup that simply cloned another AI startup
> you can't just fork a project, make, a trivial one-line change, call it a derivative work, and replace the license

yes you can

"You may reproduce and distribute copies of the Work or Derivative Works thereof in any medium, with or without modifications"

spencerchubb··on TinyJS – Shorten JavaScript QuerySelect with $ and $$
how would you add listeners if you were doing an html string? the first approach allows you to do something like this

``` const myButton = button({ onclick: e => console.log(e), textContent: "Click me" }) ```

spencerchubb··on YC criticized for backing AI startup that simply cloned another AI startup
lol maybe verify your claim before using such a snarky, arrogant tone

"You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such Derivative Works as a whole, provided Your use, reproduction, and distribution of the Work otherwise complies with the conditions stated in this License."

to be clear, Apache 2.0 requires you to give a copy of the license in a derivative, but does not require you to apply Apache 2.0 to a derivative. the term you are probably looking for is copyleft, which Apache 2.0 definitely does not have

spencerchubb··on YC criticized for backing AI startup that simply cloned another AI startup
It's perfectly legal to relicense Apache 2.0 license
spencerchubb··on YC criticized for backing AI startup that simply cloned another AI startup
Pear never advertised that they built everything themselves. They were very open about the fork
spencerchubb··on YC criticized for backing AI startup that simply cloned another AI startup
Your first point is not true. If Continue wanted a copyleft license, they would have done so. Continue basically said they are fine with people forking and changing the license
spencerchubb··on Launch HN: Inkeep (YC W23) – Copilot for Support (think Cursor for help desks)
This is the first time I've seen someone use two instances of "x for y" where x is another product and y is the actual thing they do
spencerchubb··on Visual Studio Code is designed to fracture (2022)
I can give you one data point. There is no way in hell my company would let us use cursor. Generative AI is viewed with a lot of skepticism, and we spent a long time testing out copilot before approving it for all developers to use. We use copilot because microsoft is so trusted and we use many microsoft products
spencerchubb··on Visual Studio Code is designed to fracture (2022)
What kind of set up do you use? This does not resonate at all with my experience of vs code
spencerchubb··on Ask HN: What are you working on (September 2024)?
are you designing the UI for every service by hand, or doing it programmatically somehow?
spencerchubb··on If AI seems smarter, it's thanks to smarter human trainers
Do you believe that holding a degree is dumb, or just that holding a degree is an insufficient condition for smartness? Technically what you wrote says the former
spencerchubb··on If AI seems smarter, it's thanks to smarter human trainers
> I saw plenth of questions that only had definitive answers if you bought into specific theoretical frameworks

That kind of stuff would be great to train on. As long as the answer says something like "If you abide by x framework, then y"

spencerchubb··on What I tell people new to on-call
also lived in the midwest my whole life, mix of rural and cities; this is false
spencerchubb··on What I tell people new to on-call
do you work with any women or 100% men?
spencerchubb··on I Am Tired of AI
LLMs have been improving exponentially for a few years. let's at least wait until exponential improvements slow down to make a judgement about their potential
spencerchubb··on Our Android app is frozen in carbonite
yeah I think android's policy is pretty reasonable here. if you're gonna have read/write access to everything in my google drive, you should be scrutinized pretty heavily.
spencerchubb··on Orion, our first true augmented reality glasses
why?
spencerchubb··on The Intelligence Age
Labor organizers mostly make things worse for laborers. Capitalism has brought much prosperity. "It is not from the benevolence of the butcher, the brewer, or the baker, that we expect our dinner, but from their regard to their own interest"
spencerchubb··on The Intelligence Age
To raise $1e7, you have to be able to convince VCs that you can make at least $1e9. They want 100x returns from their winners, and don't care about anything else in their portfolio.

If instead you only need $1e3 to build it, it doesn't have to make as much money. I could just fund it out of pocket

spencerchubb··on The Intelligence Age
Yes you are correct that jumps in intelligence were enabled by exponential increases in scale. That makes me more bullish on AI, not less. It suggests that we can continue exponentially scaling compute like we have done for the past few decades, and also get intelligence improvements from it.
← PreviousPage 3 of 16Next →