Timeline for Labs experiment launch: stackoverflow.ai
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 2 at 12:17 | comment | added | Ash Zade Staff | Thanks for posting. Re: model agnostic - this is so we can react to the continual and rapid model advances. In reality, we'll use a combination of models depending on the task. In its current iteration, we store convos so that we can analyze them to improve the product. For quotas, we've modelled for a lot of free usage so we have time to learn what's the best balance for us and our users. The model provider we're using provides additional moderation tooling, which we've implemented. | |
| Jul 2 at 10:37 | comment | added | tkruse | Agentic has many meanings, in this case it could mean making multiple RAG requests recursively. It doesn't need to imply other actions are possible to the ai | |
| Jul 2 at 6:09 | comment | added | Adamant | "No agentic features" is probably an advantage in this case, since it means a model that won't try to buy a dozen eggs for 30 dollars, turn off oxygen, blackmail programmers, or sell me tungsten. We don't actually need that to answer SE questions, and I frankly believe this is just one way that the "one model to rule them all" approach of AI companies is potentially detrimental. | |
| Jun 30 at 22:17 | history | edited | tkruse | CC BY-SA 4.0 | added 228 characters in body |
| Jun 30 at 22:11 | history | answered | tkruse | CC BY-SA 4.0 |