odo1242

3 ms·

1.
by odo1242 5d ago
I feel like my point is being over extrapolated a bit. All I was saying is that people in 2020 believed AI companies had much more of a moat then they actually do - that other companies would struggle to obtain enough data (y’know, download
2.
by odo1242 5d ago
They probably weren't expecting AI foundation models / model research to be quite as fungible as they are
3.
by odo1242 6d ago
Most of the GPU cost is in the GPUs themselves (and in the space and maintenance costs of the building). Electricity is a small fraction, and it's not like datacenters are just going to shut down their servers when they're not in
4.
by odo1242 7d ago
I mean, it’s the same on every browser except Safari
5.
by odo1242 9d ago
Yea, reduce is most useful as a parallel operation, like in MapReduce (Or, at minimum, when the reduction operation is commutative)
6.
by odo1242 9d ago
But the built in AI model in Chrome is 4GB of RAM, 4GB of disk space, and is still only used in a couple places
7.
by odo1242 9d ago
The number of options has to be small and bounded. The query planning is more of a search/optimization problem than a classification problem since the number of options increases wildly based on query size.
8.
by odo1242 15d ago
Well, that’s basically just the California law everyone hated
9.
by odo1242 15d ago
Ah
10.
by odo1242 15d ago
> A headwind makes it harder to advance in the direction you're going nit: except in aviation
11.
by odo1242 19d ago
It seems like the main disadvantage would be that you have to load all the autocomplete data / model weights on the client-side and your webpage might be CPU/memory limited
12.
by odo1242 19d ago
The client is doing the work in both cases
13.
by odo1242 19d ago
Well, there was CoinHive which did this exact thing 6-7 years ago, but that system got abused a bit much
14.
by odo1242 20d ago
AIs achieving legal personhood isn’t going to do much. You can’t just put an AI behind bars
15.
by odo1242 21d ago
Government funding must be approved by Congress. If Congress already approved funding the NSF, then it has to be Congress that removes the funding.
16.
by odo1242 21d ago
Second order code loss: make a change, realize you screwed everything up, unable to revert
17.
by odo1242 21d ago
Fair, but like, why??
18.
by odo1242 21d ago
I mean, versioning is basically free from a technical / storage / time spent standpoint, and my experience says that if you don’t do it you likely will lose your code at some point
19.
by odo1242 22d ago
On the contrary, it lets you MITM encrypted communications by swapping the website's original certificate for the "well trusted TLS cert"
20.
by odo1242 22d ago
ml5.js has a similar license: https://github.com/ml5js/ml5-library?tab=License-1-ov-file Their license is basically "use freely unless our community decides you violated the code of conduct, for three years"
21.
by odo1242 23d ago
I'm just going to borrow a sibling comment: > afiori: JavaScript's execution model uses cooperative concurrency to avoid a tons of data races, preemptive non-parallel concurrency would make all web development incredibly harder
22.
by odo1242 23d ago
No, I'm just saying that cooperative multitasking is there for good reason and isn't necessarily a bad idea.
23.
by odo1242 23d ago
Update: this is because ChatGPT was cheating and reading my conversation history. The closest I can get on a standard LLM is golffizbuz (10 characters)
24.
by odo1242 23d ago
My attempt: > Prompt: mkfzbuzgolf > LLM Used: Default ChatGPT free LLM > Output: for i in range(1,101):print("Fizz"*(i%3<1)+"Buzz"*(i%5<1)or i) So, 11 prompt characters and 62 output. ——- Did a few more tr
25.
by odo1242 23d ago
Cooperative multitasking isn’t universally a bad idea (in fact, Rust supports it)
26.
by odo1242 23d ago
It’s the same company that bought MuseScore, they basically plan on monetizing it through plugins, audio samples, or optional cloud services (like MuseScore)
27.
by odo1242 23d ago
Unironically, this is a browser extension I wanna make lol
28.
by odo1242 23d ago
Nice, I see how that affects the cost benefit analysis for sure
29.
by odo1242 23d ago
So, basically, sure, but I disagree with the hypothetical in the first place
30.
by odo1242 23d ago
I mean it could also get dropped yea, sure. This is just a neutral discussion of the costs, I'm not making an argument for anything. (As for data, the third hard disk isn't actually bought yet and is meant to represent an offsite

More ›