━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ MY THOUGHTS ON ARTIFICIAL INTELLIGENCE (LLMS IE. CHAT-GPT, GOOGLE BARD) anya ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Table of Contents ───────────────── Disclaimer Things I like about LLMs Things I dislike about LLMs .. Most of them aren’t FOSS (Complaints about non-FOSS LLMs) .. General Complaints Things I am looking foward Disclaimer ══════════ I am not a professional on AI or LLMs, I’m just a guy on the internet sharing his opinions. If I got something wrong on this post please contact me to let me know. Things I like about LLMs ════════════════════════ • Makes tedious work much easier to do especially in a professional environment. • Great for Tasks that don’t require Critical Thinking ie. Email templates, Essay Outlining, Summarizing things, Adding Filler words to your project (Really good for students). • Gives me better answers than a search engine for most (trival) questions and the answers are simpler, usually. Things I dislike about LLMs ═══════════════════════════ Most of them aren’t FOSS (Complaints about non-FOSS LLMs) ───────────────────────────────────────────────────────── • They are online only, making it easy violate privacy and worsen anonymity. They are probably recording everything you do on the site to feed it back into their Dataset. • Most LLMs are made by big corporations ie. Chat-GPT(Funded by Microsoft), Google Bard (Google), which causes monopolies. • Closed Datasets (The data might be biased), We also don’t know where they are getting this data and if it is ethical or not. Is the dataset that Chat-GPT was trained with just public information on the internet or did OpenAI also get shady info from Outlook users emails (they are funded heavily by Microsoft). We will never know because the Dataset is not Open. General Complaints ────────────────── • Hallucinations (Sometimes LLM’s Hallucinate and make up things that aren’t real) Things I am looking foward ══════════════════════════ • An open source LLM • FOSS offline LLMs with open datasets