ipsum2

34 ms·

571.
by ipsum2 3y ago
I haven't tested asking it explicitly to search, but it does incorporate answers that are very recent and unlikely to be in it's training dataset
572.
by ipsum2 3y ago
While technically true, you could say that about all riddles. See GPT-4's explanation here, which intuits the rule and answers the riddle correctly. https://chat.openai.com/share/b1452950-b493-4e27-b097-e64f21...
573.
by ipsum2 3y ago
Great prompt. GPT-4 was able to answer this, but Gemini Ultra was not: Jake likes coke but not pepsi. He likes corn but not popcorn, and he likes pens but not pencils. Will Jake like salmon or cheese?
574.
by ipsum2 3y ago
Note that Gemini pulled the answer off the Internet, while GPT-4 didn't. The answer can easily be found via Google search. Changing up the question a little, I reversed it and asked Ultra and it was unable to answer: Jake likes coke bu
575.
by ipsum2 3y ago
In the 20 minutes of experimentation, I'm really impressed with the quality of Bard Advanced (Gemini Ultra). The results are as good as GPT-4, and in some cases is better. So far: pros: - better at translation (tried Chinese and Japane
576.
by ipsum2 3y ago
The tl;dr seems to be: Tell a LLM to create pairs of questions and answers based off of a document, and fine-tune on that data. Does the model answer questions from the article that weren't generated in advance?
577.
by ipsum2 3y ago
I built this for myself, the main bottleneck is TTS prices are too high for an entire book, and open source ones aren't good enough yet.
578.
by ipsum2 3y ago
I saw your project at Open Sauce last year, happy to see the official release.
579.
by ipsum2 3y ago
MIT has an atrocious ML department, all the papers I've read from them have been third-rate. Mid-tier colleges in Asia have better researched papers.
580.
by ipsum2 3y ago
This isn't for tabular data though? Time series transformers have been around for a few years now, see Transformers in Time Series: A Survey https://arxiv.org/abs/2202.07125 and Are Transformers Effective for Ti
581.
by ipsum2 3y ago
I don't know why you were down voted for sharing your opinion on bard. I agree with you that bard is significantly worse than gpt 3.5.
582.
by ipsum2 3y ago
Context: OpenAI was "caught" using Matryoshka embeddings in their new release. They apologized and added references to the paper in their release notes.
583.
by ipsum2 3y ago
Chrome's PDF reader reports the file size as disappointly 200.00 × 200.00 in (square)
584.
by ipsum2 3y ago
On mobile you click, on desktop you drag and drop.
585.
by ipsum2 3y ago
Side loading apps on the quest is pretty simple using the aptly named SideQuest, it only requires a one time 15 minute set up. Definitely a lot easier than trying to do this on an iPhone/iPad/iQuest.
586.
by ipsum2 3y ago
That comparison doesn't make much sense, since Ballmer made his money by holding on to Microsoft stock after he left.
587.
by ipsum2 3y ago
TIL radio buttons were named after the physical buttons used on older radios to select preset stations[0]. IMO they look nothing like the actual buttons on radios. 0: https://en.wikipedia.org/wiki/Radio_button
588.
by ipsum2 3y ago
Mojo is a closed source language that will never reach mainstream adoption among ML engineers and scientists.
589.
by ipsum2 3y ago
See reply to sibling comment.
590.
by ipsum2 3y ago
The model being compared on the LLM Arena does not use ChatGPT4, it uses GPT4 API, which does not have access to the internet by default.
591.
by ipsum2 3y ago
It's not a valid comparison. Bard uses Google (ie the Internet) to answer things, while GPT4 doesn't. So bard can answer things like, "what's the weather in SF today" or "who won the basketball game last night&
592.
by ipsum2 3y ago
Did you see that the person who wrote the article also created https://www.vanillabreeze.dev/ to remove dependencies from Tailwind?
593.
by ipsum2 3y ago
Incorrect, most of the libraries can run on edge, they're just C++.
594.
by ipsum2 3y ago
It doesn't require it in theory, but in practice its required bc CPUs are too slow at fine-tuning and computing embeddings.
595.
by ipsum2 3y ago
Hydrogen balloons are fun to pop.
596.
by ipsum2 3y ago
Llama2 wasn't trained on code, so it's no surprise that you're having trouble generating SQL.
597.
by ipsum2 3y ago
Hetzner is in Germany, not sure about France.
598.
by ipsum2 3y ago
Google, not content with existing image formats that support 10-bit like HEIC (used by Apple) and AVIF (based off of AV1, a codec that Google helped design and is better than JPEG), they decided in all their wisdom to make Ultra HDR for And
599.
by ipsum2 3y ago
You can run llama models on a personal computer, even though it was trained on >10,000 GPUs.
600.
by ipsum2 3y ago
Do normal proofs for Olympiad level geometry questions require 150+ steps like the one that was generated here? It seems inelegant.

More ›