wsgeorge

24 ms·

211.
by wsgeorge 4y ago
This is exciting! The open source LLM space is about a year behind OpenAI's SOTA, so any step towards closing that gap is welcome It's still a bit of a shame Meta hasn't considered a more permissive license for LLaMA
212.
by wsgeorge 4y ago
> Is there something called passion and do everyone have it? Yes, there is. No, not everyone has it. I stopped looking for passion a long time ago. What drives me more is fear of failure/disappointment. :|
213.
by wsgeorge 4y ago
Abstractions as a concept. Can't unsee them now, 14 years on.
214.
by wsgeorge 4y ago
Because it would be against their TOS, and things could look ugly, legally.
215.
by wsgeorge 4y ago
Depends on what you mean by AI blockers. If they identify themselves honestly, yes. https://platform.openai.com/docs/plugins/bot
216.
Ideas Will Be King(wsgeorge.medium.com)
1 points by wsgeorge 4y ago|0 comments
217.
by wsgeorge 4y ago
Group-think is a pro-social strategy. Getting comfortable with being "the odd one" is a good first step. Although it can be debilitating in excess, cultivating a critical attitude (which means frequently re-examining assumptions a
218.
The Claude Glass, also known as the “18th-century Instagram filter“(twitter.com)
2 points by wsgeorge 4y ago|0 comments
219.
by wsgeorge 4y ago
How are they ordered? There's a few open source LLMs that aren't in this index, and it's not clear who the providers are supposed to be (AFAIK, alpaca isn't "provided" by any one party)
220.
by wsgeorge 4y ago
The programming language is an intermediary we humans and the machine can understand. It's better from an audit/explainability perspective.
221.
by wsgeorge 4y ago
This brings back memories from Obama's first campaign. My parents joked that with the kind of influence the US has over us, every country should have a say in who leads them! "Global Voting" is a poor solution to that problem
222.
by wsgeorge 4y ago
The only reason I stopped using Google Search was because it often interrupted my requests with a CAPTCHA asking me to prove I was human, which I always found insulting. Besides that, I've never had a problem with the quality of result
223.
by wsgeorge 4y ago
> but I think this one is just different. It might change the world really drastically. FWIW crypto did change the world in a number of ways: 1. CBDCs actively being explored. 2. Increased concern about the environmental cost of compute.
224.
by wsgeorge 4y ago
> Personally, I think it's an interesting choice, but it may not be for everyone. That's the point, I think. If I were in their shoes, I will also start with a slow rollout that limits the number of users to better understand h
225.
by wsgeorge 4y ago
You're right about that. I bet the algolia search is enough for HN's requirements. https://hn.algolia.com/?dateRange=pastYear&page=0&prefix=tru...
226.
by wsgeorge 4y ago
No. This too shall pass. :)
227.
by wsgeorge 4y ago
> But why would you supply untrusted input to AI code completion or your private AI assistant? I believe that's the OP's point. Copy-pasting some text from a site into a private LLM is very easy (and fun) to do when exploring t
228.
by wsgeorge 4y ago
> it didn't impact any deployment plans. With a market this fresh and heated, NOTHING will impact deployment plans except for backlash when things go awry after deployment. This space is going to be even more interesting than the
229.
by wsgeorge 4y ago
I'm not really bothered by the predicted fake news and spam deluges: I've seen enough of that already to know that regular social media use is not for me, and I'm generally picky about the topics I consume, and who I get them
230.
by wsgeorge 4y ago
It's little things like this, and big things like llama.cpp that convince me of Gerganov's superpowers :)
231.
by wsgeorge 4y ago
> In the following example, let’s imagine a new AI assistant, Bong I laughed too hard at this! The fact that LLMs deployed en masse open up new security threats - socially engineering AIs to act maliciously - is both exciting and terrify
232.
by wsgeorge 4y ago
Reminds me of this project submitted yesterday [0]. I'm trying hard to keep up with the pace of projects and papers being announced. This is all very exciting! [0] https://zero123.cs.columbia.edu/
233.
by wsgeorge 4y ago
On first look, this reminds me of Google Wave...
234.
by wsgeorge 4y ago
> For translation, you are better off starting from a model specifically trained for that purpose Yes, but wasn't the whole point of the recent LLM research to show that you didn't need to fine-tine for a specific task?
235.
by wsgeorge 4y ago
It was more fun to see GPT-4 acknowledged in the GPT-4 technical report [0] OpenAI published. [0] https://arxiv.org/abs/2303.08774
236.
by wsgeorge 4y ago
I chuckled at one of them. They're not that bad.
237.
by wsgeorge 4y ago
You'd imagine that we would have move on from this by this time but no. Also not available in my country. :')
238.
by wsgeorge 4y ago
The list of features look impressive, honestly speaking. Image completion really got me. This is what it looks like when a company is serious about commericalizing generative AI. If the landing page isn't just cherrypicked marketing hy
239.
Consumer AI Is Ripe for Centralization(medium.com)
1 points by wsgeorge 4y ago|0 comments
240.
by wsgeorge 4y ago
My hunch is that following an instruction is a subset of text completion. > How are you? The most likely next word might be computed to be "I". > How are you? I Then, what's the word that's most likely to follow &q

More ›