PinnedMike YounginGeek CultureMy Accidental Path to 1,000 Dev.to FollowersMy path to almost-accidental popularity on Dev.toJun 25, 20232Jun 25, 20232
Mike YounginArtificial Intelligence in Plain EnglishAI agents can collude using hidden messagesLLMs can hide their real messages in unsuspicious chit-chat1d ago1d ago
Mike Young🧠 Training on code improves LLM performance on non-coding tasksAdding code to your training data makes your LLM better at non-coding tasks tooAug 30Aug 30
Mike YoungLLMs can speak in JPEGBy studying “secret” coded messages (compressed JPEGs), LLMs can eventually learn to write their own.Aug 22Aug 22
Mike YoungThe GPT store is stupid and deadThere’s no moat. All prompts can be extracted, so all prompts are public.Aug 3Aug 3
Mike YoungCan you trick LLMs into giving dangerous info by phrasing questions in past tense?Doctors hate him! Get 8,000% more “dangerous” answers with this one weird trick!Jul 20Jul 20
Mike Young10–30% of today’s research is written by AIChatGPT is used for tons of science writing, especially in certain fields and countriesJul 12Jul 12