dcastm

8 ms·

61.
Ask HN: Is anyone using GPTs regularly?
1 points by dcastm 3y ago|1 comments
62.
by dcastm 3y ago
If you think objectively about this issue, it's hard to see how they can get a positive ROI out of this. Simon does a great job highlighting why this doesn't make sense from the technical and business standpoints: https:/&#x
63.
by dcastm 3y ago
People often get mixed up about what causes what in these situations: We might chat about all sorts of random things, but many times there's something that happened before, let's call it event X, that leads us to talk about a cert
64.
by dcastm 3y ago
Why do you need a chatbot? I'm genuinely wondering about it. It feels that you could use the same 3/4 questions and then send the results to the user. Maybe the "smart" part of it would be giving recommendations based on
65.
by dcastm 3y ago
I think we'll start seeing a lot more services like https://www.together.ai soon. Having open-weight models better than gpt-3.5 will drive a lot of competition on the LLM infra.
66.
by dcastm 3y ago
If you take input tokens in consideration is more like 5.25 eur vs. 1.5 eur / million tokens overall. Mistral-small seems to be the most direct competitor to gpt-3.5 and it’s cheaper (1.2 eur / million tokens) Note: I’m assuming e
67.
by dcastm 3y ago
GPT-3.5 is probably the most popular model used in applications (due to price point vs. GPT-4, and quality of results vs. open-weight models). So I guess they're trying to say now it's a no-brainer to switch to open-weight models.
68.
Finetuning won't add new knowledge to your model(dylancastillo.co)
7 points by dcastm 3y ago|0 comments
69.
by dcastm 3y ago
Very interesting. I didn’t know about it. Thanks!
70.
by dcastm 3y ago
I guess that at a big enough scale it starts making sense. But not sure what “big” really is. Pgvector seems to perform very poorly compared to Qdrant: https://nirantk.com/writing/pgvector-vs-qdrant/
71.
by dcastm 3y ago
Me too! I started programming because of Game Maker. I lost all the games I did. I wished I had saved them somewhere.
72.
Show HN: Write real estate listings using AI(lystinct.com)
1 points by dcastm 3y ago|0 comments
73.
by dcastm 3y ago
I wonder how much better is this, compared to taking the average ( or some other aggregation) of embeddings with a smaller context length. Has anyone done a similar comparison?
74.
by dcastm 3y ago
You were probably using the chat version which has been moderated, and hhh used the base version.
75.
OpenAI Cookbook(cookbook.openai.com)
4 points by dcastm 3y ago|0 comments
76.
by dcastm 3y ago
For those curious about it, ChromaDB uses all-MiniLM-L6-v2[0] from Sentence Transformers[1] by default. [0] https://docs.trychroma.com/embeddings#default-all-minilm-l6-... [1] https://www.sbert.net/docs&#x2F
77.
by dcastm 3y ago
It’s under Growth > Files for those struggling to find the button
78.
by dcastm 3y ago
Super cool. Here the viz-related prompts (generation, editing, etc), for those interested: https://github.com/microsoft/lida/tree/main/lida/components/... I built a tool that lets you use GPT t
79.
by dcastm 3y ago
Now they can hire 4 more engineers :P
80.
by dcastm 3y ago
> This blog, and the RSS feed will slowly grow into read-only automation that will publish to all of my channels. WUPHF?
81.
by dcastm 3y ago
Amazing! Gonna check it out. Thank you!
82.
by dcastm 3y ago
Awesome work! I love the simplicity of htmx. It’s one of those tools that makes me feel super productive. Will there ever be a “htmx“ for building mobile apps?
83.
by dcastm 3y ago
Check Cohere's embeddings of Wikipedia: https://txt.cohere.com/embedding-archives-wikipedia/
84.
by dcastm 3y ago
The article is referring to the problem of having a limited context length in LLMs. That is you can only pass X tokens in the prompt. For example, let’s say you have a prompt that lets you answer questions about a book. If the book is long
85.
by dcastm 3y ago
Taking this to its logical conclusion, we should all be speaking Esperanto. My guess is that there's a combination of network effects (i.e., 1.4 billion people already use it), cultural identity, and inertia.
86.
by dcastm 3y ago
You could close slack and open it a couple of times a day. The email is an unnecessary intermediate step.
87.
by dcastm 3y ago
The part I hate the most about Slack is that when you join a new Workspace, the default is to send you an email with notifications if you’re inactive more than 15 minutes. And to turn it off, you got to go into a kind of hidden configuratio
88.
by dcastm 3y ago
The title is wrong. There's no effect here, just correlation. It's pointless to draw any conclusions from this study. If you think otherwise, check https://www.tylervigen.com/spurious-correlations
89.
by dcastm 3y ago
This works well for simple use cases, but quickly breaks down when faced with real-life scenarios. Database schemas might contain hundreds of tables, and they may not always have intuitive names. Also, the relationships between the tables a
90.
by dcastm 3y ago
Interesting. But I believe you're be wrong about Show HN (as others have pointed out). Show HN makes for an effective way to get to the front page, based on three reasons: 1. You don't need that many votes to make it to the front

More ›