Jose
jsameijeiras.bsky.social
Jose
@jsameijeiras.bsky.social
Economics | Data Science | AI | Phylosophy | Art
Yes totally but my only point is that we cannot pool post soviet countries now members of the EU having a dem backlash with Venezuela or Philippines.

Spec, cause this might give the impression that things are going much worse than they are.
November 6, 2024 at 8:16 PM
Okay but are you trying to argue that's the same as Venezuela or Philippines?

How does the judicial system compare to the one 40 years ago?
Didn't individual freedom has increased in both those countries in recent hist?
The EU rule of law debate doesnt only apply to those countries...
November 6, 2024 at 3:02 PM
🤔🤔🤔
November 6, 2024 at 2:45 PM
Not sure why you're putting two EU member, with counterbalance power from Brussels, in the same pool as Venezuela, Philippines and Turkey.
November 6, 2024 at 1:51 PM
I'll check it out!
November 4, 2024 at 8:48 PM
Noo! I've never use Positron, I tried Cursor and the likes but only with Python!
November 4, 2024 at 8:39 PM
I couldn't check, but I hope so, otherwise...
November 4, 2024 at 8:26 PM
Agree - and that's why the SQL bit you were mentioning is interesting, I wonder how you orchestrate it.

In any case great post and thank you for sharing!
November 4, 2024 at 8:24 PM
Did someone really thought an alternative was possible? 🤔😅
November 4, 2024 at 8:23 PM
Also, the fact that are probabilistic means that they will be wrong on the less representative topics, outliers etc... Therefore this can mean that the 5% that is wrong is actually less significant, and probably more debatable?
November 4, 2024 at 8:18 PM
I am very confused by this argument. Doesn't this depend more on the critical mind of the reader/searcher and not the accuracy of the answer.

Also, 95% of the answers right means that for those that don't fact check, are 95% of the times right.
November 4, 2024 at 8:17 PM
Yes! We are talking about a database of legal decisions. I was experienced using it, and instead of using filters the GPT allows me to write down a request in NL. This is good, because the GPT immediately does connections and queries the API using synonyms or related concepts, that maybe I wouldn't
November 4, 2024 at 8:14 PM
Yes - fair! But still, my little using a GPT to query the European Commission database on merger control, is that it would give me decent answers if you give the proper schema. So I wonder how much it can scale, and how much is OpenAI doing the right work...
November 4, 2024 at 8:10 PM
Gracias!
November 4, 2024 at 6:52 PM
Great tutorial milos! Btw I see you use visual studio instead of Rstudio, any specific reason or it's just not change editor when you are working on Python.
November 4, 2024 at 4:18 PM
So you don't believe that the LLMs could become proper agents to handle the necessarily steps to query data if given the right access to tools, like in your case SQL capabilities?
November 4, 2024 at 4:02 PM
Can we make chatbot interface to interact with it. I am thinking of giving a chatgpt acces to your dataset in HuggingFace and then anyone could ask a question and get the INE data?

Also, only 500mb ?!?
November 4, 2024 at 3:23 PM
It's so nice! I guess there are some limitations for defense and critical infrastructure code.
November 4, 2024 at 11:24 AM
Reposted by Jose
My starter packs include:
- AI and data people go.bsky.app/7D4NApV
- Tech managers go.bsky.app/J3W39y1
- ML/AI researchers go.bsky.app/gPKeHz
- Hugging face people go.bsky.app/PTmCYsU
November 3, 2024 at 7:33 PM
Gracias por compartir Pelayo! Podrías añadirme al de AI and data people ☺️
November 4, 2024 at 9:17 AM
Noted! Thank you Randy!
November 3, 2024 at 11:25 PM