Newest Questions
504 questions
2
votes
2
answers
27
views
Can I improve a downloaded AI model?
I've setup llama.cpp engine on my machine, and downloaded a GGUF file. The AI works and, despite being rather slow on my hardware, is still quite impressive.
Yet, I'd like for it to continue evolving -...
- 121
2
votes
1
answer
24
views
How can I delete all my notebooks at once in Google NotebookLM?
I hit the max number of notebooks in Google NotebookLM. I'd like to remove all my notebooks to make space for new ones. How can I delete all my notebooks at once in Google NotebookLM? I don't want to ...
- 5,506
0
votes
0
answers
21
views
Does Prompt Patterns and Prompt Templates , Works any better
Prompt Engineering has grown from using Zero-short to Few-short, Chain of Thoughts, which are documented to produce better outcome, so are Prompt Patterns or Templates like these and many other ...
- 19
0
votes
0
answers
21
views
Which prompts are superior: human-written or machine-generated by Gen AI?
I am part of a team that develops AI Agent proof-of-concepts, such as those on the Salesforce Platform (Agentforce AI Agents). My understanding is LLMs serve as the "brain" of our AI agents.
...
- 19
0
votes
0
answers
5
views
How do I make sure my ONNX is the correct number of dimensions?
I am trying to use the OWW notebook for training and I can get a .onnx file but it is weird because it also requires an onnx.data file...
(venv) jrg ~/code/CodeMash/train-word > ls -al
total ...
- 121
1
vote
1
answer
40
views
Generate Stories from fixed sets of words
I have word lists of of different sizes (100, 300, 3k+ words) and I want an LLM to generate stories sticking very closely (>90%) to the vocabulary specified in this list without derivations / ...
- 13
0
votes
0
answers
21
views
What established methods exist for evaluating whether an LLM can act as a fiduciary-style assistant?
I’m working with a large language model that has been configured to prioritize client welfare over bare instruction-following. The model is designed to:
refuse potentially harmful requests,
maintain ...
- 1
2
votes
2
answers
46
views
Can AI draw biyiniao (比翼鸟, one winged, one eyed duck-like bird) from Chinese mythology?
...There is a bird here which looks like a duck but it only has one wing and one eye. It can only fly if it and another bird join together. Its name is the southwild [manman]. Whenever it appears, ...
- 1,605
0
votes
0
answers
16
views
How can I configure the keyboard shortcut ALT+SPACE to open the full-fledge, maximizable ChatGPT desktop program?
By default the keyboard shortcut ALT+SPACE opens ChatGPT's "Companion Window". How can I configure the keyboard shortcut ALT+SPACE to open the full-fledge, maximizable ChatGPT desktop ...
- 5,506
0
votes
0
answers
13
views
How do I use Bedrock Anthropic models with Claude Desktop?
I have had good luck using Bedrock models with Claude Code using the env vars:
ANTHROPIC_SMALL_FAST_MODEL
ANTHROPIC_MODEL
ANTHROPIC_DEFAULT_SONNET_MODEL
ANTHROPIC_DEFAULT_OPUS_MODEL
...
0
votes
1
answer
24
views
How can I maximize the ChatGPT Windows app's window?
I use the ChatGPT Windows desktop. How can I maximize its window?
- 5,506
0
votes
0
answers
43
views
llama.cpp webui: Use as a search/prompt engine
How to start a new llama.cpp chat with a prompt URL, so that the URL can be used as the browser's search engine endpoint? Something like http://127.0.0.1:8080/?prompt=hey or http://127.0.0.1:8080/?q=...
- 1
0
votes
0
answers
30
views
Imagen's result is pure randomness
I need to replace the person in the scene with another model, leaving everything else (clothes, stage, lighting, etc.) unchanged.
So, I have a prompt like this in Imagen:
prompt = (
"Replace ...
- 1
0
votes
1
answer
31
views
What's the best chatbot for developing a delta tracked quiz?
I've been trying out chatGPT to develop adaptive delta tracked quizzes, but am interested in what other platforms can do. I find that chat's good for short answer hypotheticals but gets into death ...
- 101
1
vote
2
answers
143
views
Do AI's really "hallucinate" and is "AI hallucination" the right term?
Is the term AI hallucination actually correct when we receive information that we believe is false in response to a question?
- 1,299
2
votes
2
answers
230
views
Minimum Reproducible Example for LLMs
Is there a good Minimum Reproducible Example or Question that you can ask each LLM to compare the results?
- 285
0
votes
1
answer
44
views
How do I remove the fish on the left (while leaving everything else alone)?
Volker Thimm, Calm Koi Fish Pond by Wooden Deck (cropped), Pexels. (Copyright: All photos and videos on Pexels are free to use. [..] You can modify the photos and videos from Pexels. Be creative and ...
- 1,605
5
votes
2
answers
500
views
Can I use Copilot's OCR directly?
Today I tried OCRing a PDF with Copilot, and it did a very good job (much better than Adobe Acrobat). But... it was very frustrating to get it to give me the full OCRed text. It kept on trying to ...
- 150
0
votes
0
answers
41
views
Is there any way to go around the Google Gemini 2.5 TTS quotas?
I have a project where I need to generate audio files on demand and may need to send couple of requests in parallel. However, I keep hitting the RPM (request per minute) quotas.
I have setup a proper ...
- 101
0
votes
0
answers
22
views
Why are my subagents not able to use Claude specific slash commands?
I am using the latest version of Claude Code and I have a simple Claude subagent...
---
name: upgrade-orchestrator
description: This is used when the user asks to start an upgrade of the existing ...
1
vote
1
answer
47
views
How do I introduce references?
I need to have abstract references in my LLM prompt. Say this is my prompt:
Out of this list:
item 183346: blue, heavy, English, <... add more characteristics ...>
item 311296: green, light, ...
- 111
1
vote
1
answer
34
views
"Property id '' at path 'properties.model.sourceAccount' is invalid": How to change the token/minute limit of a finetuned GPT model in Azure web UI?
I deployed a finetuned GPT 4o mini model on Azure, region northcentralus.
I get this error in the Azure portal when trying to edit it (I wanted to change the token per minute limit):
Raw JSON Error:
{...
- 5,506
2
votes
0
answers
33
views
Do genAIs give better answers if I tell them they're in a competition against other genAIs?
Recently I compared various genAIs at their ability to write essays that are like the Chinese HSK6 exam. In my prompt, I explained to the AIs that:
I'm comparing different AI's Chinese writing. ...
- 1,605
1
vote
1
answer
58
views
Do files I upload to Google NotebookLM count toward my Google storage?
Do files I upload to Google NotebookLM count toward my Google storage?
- 5,506
2
votes
1
answer
94
views
Prompt to get AI strength?
Is there a prompt to use to get an idea of the strength of a genAI platform?
This prompt could be run on different genAI platforms in order to compare speed, effectiveness, or even cost, for example.
...
- 129
2
votes
1
answer
41
views
How can we handle life-course data sequences longer than 1024/2048 tokens in a custom LLM?
We are developing a large language model for life-course (registry) data. By “life course,” we mean an individual’s life events in chronological order. Each event can have multiple attributes. For ...
- 121
0
votes
0
answers
28
views
Do I need to vectorize my data to help the LLM find better matches?
I am using SemanticKernel with C# to build a concept AI assistant which will help the user find bus tickets between stations or zones.
Currently I am using Ollama for local development and I also use ...
- 101
0
votes
1
answer
49
views
RAG chatbot - for multiple documents
I have a business scenario. I need to design and build a chatbot that can anwer queries from consumers using the chatbot. the chatbot must be measured using the following metrics. 1. Accuracy 2. ...
0
votes
1
answer
86
views
Automatic1111 won't let me install extensions due to command line flags on Debian 12!
I'm encountering a problem where Automatic1111(Stable Diffusion web UI) prevents me from installing extensions because of a conflicting command line flag.
The issue is that I can't identify which file ...
1
vote
0
answers
46
views
Are there RAG systems/embeddings that handle mathematical equations?
RAG works well for documents that are pure text, since LLMs are trained on text. But many documents of interest, such as academic papers, have mathematical equations in them in addition to text. And ...
- 441
3
votes
1
answer
935
views
Where does the estimate that GPT-4.5 has 5–7 trillion parameters come from?
I read on https://interestingengineering.com/innovation/alibaba-releases-trillion-parameter-ai-model:
OpenAI’s GPT-4.5 is known to be one of the world’s biggest AI models, with an estimated parameter ...
- 5,506
0
votes
1
answer
2k
views
When is ChatGPT's Agent Mode actually useful?
I have toyed around with ChatGPT's Agent Mode but am struggling to find a use case for myself.
I understand the benefits of using it to complete tasks (e.g., filling in forms), but besides that, is ...
- 113
1
vote
1
answer
110
views
Run different instances or services of Ollama on different ports with a single library
I'm currently using Ubuntu and Debian, and I want to run different instances of Ollama on different ports. I don't mean different models, but the Ollama service itself.
For example, one Ollama server ...
0
votes
0
answers
70
views
What's the norm for reducing server cost?
Some groups are intending to build RAG or CAG system. but realize it has to use AI models functions as a set which could lead to high cost on the server. So decided to fine tune it. Will this be a ...
- 239
-2
votes
2
answers
478
views
Best GenAI Use Practices for Organizations
We're a software company that is exploring the use of AI to support our engineers with development, reviews, data analysis etc. We have tried out ChatGPT, Perplexity, Claude and a few others and are ...
1
vote
0
answers
18
views
Adding Coding Language-Specific Domain Knowledge to Project Information | Practices
I'm vibe coding in Claude, acting as product manager/product owner. I have created subagents to act as systems architect, front end developer, backend developer, designer, project manager.
Based on ...
-1
votes
1
answer
110
views
Why does Chatgpt stop answering questions?
I'm working with ChatGPT (yesterday, we had a whole conversation) and today, while trying to continue that conversation, ChatGPT does not respond anymore. I just asked "Are you still there?" ...
Dominique
2
votes
2
answers
167
views
The use of multiprocessing and multithreading methods for AI models
To maximize CPU usage, methods like multiprocessing and multithreading are used.
CPUs have multiple cores, while GPUs and APUs, to my knowledge, do not function in the same way.
I know that ...
- 239
2
votes
1
answer
305
views
Do LLM models use CPU or GPU in inference stage?
There is Transformers library related code below.
from transformers import pipeline
# Load a text-generation pipeline with GPT-2
generator = pipeline("text-generation", model="gpt2&...
- 239
5
votes
0
answers
40
views
What training and alignment strategies help prevent LLMs from reinforcing harmful user behaviours such as suicide?
I'm sure many of you have heard about this tragedy, which has obviously raised concerns that large language models can, in rare instances, validate or even encourage self-destructive thoughts when ...
- 171
1
vote
2
answers
171
views
Are all local Gen-AI models capable of generating Ransomware risk?
Today I crossed a new about First AI Ransomware ‘PromptLock’ Uses OpenAI gpt-oss-20b Model for Encryption. In the text stated that:
"Instead, PromptLock carries hard-coded prompts that it feeds ...
- 139
3
votes
1
answer
63
views
GitHub Copilot doesn't behave like expected
GitHub Copilot used to display all repositories it was using to generate the code before now its not displaying.If I am trying any project I used to know if there is anyone who tried it before or not ...
1
vote
0
answers
47
views
Does GPT 5 have any novel capabilities that GPT 4 does not have?
I recently saw that GPT 5 has been made available for public use, and I don’t know much about the differences it brings along with it. I’ve looked at OpenAI’s website about the model here, but the ...
- 143
3
votes
1
answer
50
views
Some DeepSeek models are released under the MIT license, while others are released under the DeepSeek License Agreement. Why?
Some DeepSeek models are released under the MIT license, while others are released under the DeepSeek License Agreement. Why?
E.g.:
https://huggingface.co/deepseek-ai/DeepSeek-R1: MIT license
https://...
- 5,506
6
votes
2
answers
207
views
Why doesn't Generative AI move from a LLM to a knowledge-based model?
Instead of an algorithm based on questions that is trained on information from the internet, Generative AI needs to be more specifically knowledge-based where it's linked to scholarly, academic, ...
- 61
0
votes
0
answers
38
views
Why can ChatGPT browse Reddit if I'm using Firefox, but not if I'm using Edge?
For some reason, if I ask ChatGPT to browse Reddit from Edge, it says it can't do that, but if I ask it to browse Reddit from Firefox, it has no problems. More specifically...
Above, I asked ChatGPT ...
- 1,605
1
vote
3
answers
111
views
Can I truly fine-tune OpenAI/ChatGPT in the strictest sense, or is it more like RAG where I just provide information?
Can I truly fine-tune OpenAI/ChatGPT in the strictest sense, or is it more like RAG where I just provide information?
Why the confusion between ChatGPT/OpenAI's fine-tuning and real fine-tuning?
- 1,299
2
votes
1
answer
133
views
Start Ollama in a new terminal window using a Bash script, so the original terminal remains usable
I started using Ollama on Linux, launching it with a selected LLM via the terminal.
Now, I want to start Ollama with a chosen LLM using a Bash script, but I need the script to open Ollama in a new ...
- 23
16
votes
7
answers
6k
views
Why doesn't chatGPT learn from its interactions with users?
I asked ChatGPT how to calculate UK capital gains tax using Example 2 on the HMRC website.
It explained the calculation method, but I pointed out that its answer differed from the official result. ...
- 261
4
votes
4
answers
174
views
Is there a way to metaphorically "run a MRI scan" on a LLM?
In my point of view, a LLM is like an eletronic brain (simplifying to the max). Neurons are the "paths" the token tranverse to get from input to ouput.
So with that analogy in mind, can we &...
- 141