About this episode
MedPaLM 2 outperforms human doctors in 86.5% of cases Meta AI’s 1.2m Token model Megabyte Meta’s LIMA: Less Is More for Alignment Photoshop & Adobe Firefly gets even better Crazy things Job interview real-time live transcription and responses Drag GAN - Modifying images by dragging certain points Prompt tips - Knowing how the model was trained to choose your prompt engineering style (e.g. if it was trained on Q&As, then format your prompt with a Q&A). - Give the model many examples of what you expect (e.g. if you want your answer in a certain style) - Specificity vs generality. The more specific you are, the better the answer you’ll get. - Always remember that the (transformer) model is predicting the next most probably sequence of tokens or words, whenever you are creating a prompt.
Links to other companies / models / startups mentioned
Clearword meeting notes summariser
Tactiq Mentions: Model conditioning, Tokens, Prompt Engineering, Model Training, Sequence2Sequence, RLHF, model weights, hyper parameters, supervised learning, unsupervised learning, annotated data, caselaw, finetuning, model selection, LLMs, Google, Rainbow Corn, Generative AI, Emergent capabilities Book: Michio Kaku: Physics of the future (misquoted as Science of the future)
Get every episode summarized
Each time Using AI publishes, we email you a written briefing from the transcript — the topics, who appeared, and any specific claims, with the ad reads skipped.
Email me new episodesFree for 3 shows. No card needed.
Hosts & guests
No transcript yet
This episode has not been transcribed. Request it and it moves to the front of the queue.
More episodes
