Researchers have uncovered a new supply chain attack called #Slopsquatting where threat actors exploit hallucinated, non-existent package names generated by #AI coding tools like #GPT4 and #CodeLlama

These believable yet fake packages (amounting to 19.7% or 205,000 packages), recommended in test samples were found to be fakes., can be registered by attackers to distribute malicious code.

Open-source models -- like #DeepSeek and #WizardCoder -- hallucinated more frequently, at 21.7% on average, compared to the commercial ones (5.2%) like GPT 4.

We Have a Package for You! A Comprehensive Analysis of Package Hallucinations
by Code Generating LLMs (PDF) https://arxiv.org/pdf/2406.10279

💡LM studio esegue sul tuo pc chat conversazionali
LM Studio è un programma open source e gratuito ospitato su GitHub, progettato per semplificare l'installazione e l'uso di modelli LLM sul tuo computer locale

https://gomoot.com/lm-studio-esegue-sul-tuo-pc-chat-conversazionali/

#AI #IA #intelligenzaartificiale #ChatGPT #Claude #Llama #LLM #LMstudio #MPT #orca #prompt #software #STABLEdiffusion #WizardCoder

If you are interested in using #codellama or #wizardcoder with #vscode, I have a foss #vscode / #vscodium extension which allows for fairly flexible code editing using these models.

https://github.com/balisujohn/localpilot

Since it uses #text-generation-webui as a backend, it's compatible with machines with and without GPUs and doesn't require a Docker server like Fauxpilot.

This demo is local CPU inference only with a laptop i7, with the model WizardCoder python 13b.