From 30bcb8358d8365bc420efdb4d79e137dd5576fda Mon Sep 17 00:00:00 2001 From: GitHub Actions Date: Tue, 23 Jan 2024 14:36:52 +0000 Subject: [PATCH] differences for PR #434 --- 5-outlook.md | 16 ++++++++++++++++ fig/03_tensorboard.png | Bin fig/04_conv_image.png | Bin md5sum.txt | 4 ++-- reference.md | 1 + 5 files changed, 19 insertions(+), 2 deletions(-) mode change 100755 => 100644 fig/03_tensorboard.png mode change 100755 => 100644 fig/04_conv_image.png diff --git a/5-outlook.md b/5-outlook.md index 9307a460..c1aa8bc2 100644 --- a/5-outlook.md +++ b/5-outlook.md @@ -85,6 +85,22 @@ in this course. This is quite common for applied deep learning projects. It is s deep learning problem is spent on data preparation, and only 10% on modeling! ::: +::: discussion +## Large Language Models and prompt engineering +Large Language Models (LLMs) are deep learning models that are able to perform general-purpose language generation. +They are trained on large amounts of texts, such all pages of Wikipedia. +In recent years the quality of LLMs language understanding and generation has increased tremendously, and since the launch of generative chatbot ChatGPT in 2022 the power of LLMs is now appreciated by the general public. + +It is becoming more and more feasible to unleash this power in scientific research. For example, the authors of [Zheng et al. (2023)](https://doi.org/10.1021/jacs.3c05819) guided ChatGPT in the automation of extracting chemical information from a large amount of research articles. The authors did not implement a deep learning model themselves, but instead they designed the right input for ChatGPT (called a 'prompt') that would produce optimal outputs. This is called prompt engineering. A highly simplified example of such a prompt would be: "Given compounds X and Y and context Z, what are the chemical details of the reaction?" + +Developments in LLM research are moving fast, at the end of 2023 the newest ChatGPT version [could take images and sound as input](https://openai.com/blog/chatgpt-can-now-see-hear-and-speak). +In theory, this means that you can solve the Cifar-10 image classificaiton problem from the previous episode by prompt engineering, with prompts similar to "Which out of these categories: [LIST OF CATEGORIES] is depicted in the image". + +Do you agree with the following statement: + +_In a few years most machine learning problems in scientific research can be solved with prompt engineering._ +::: + ## Organising deep learning projects As you might have noticed already in this course, deep learning projects can quickly become messy. Here follow some best practices for keeping your projects organized: diff --git a/fig/03_tensorboard.png b/fig/03_tensorboard.png old mode 100755 new mode 100644 diff --git a/fig/04_conv_image.png b/fig/04_conv_image.png old mode 100755 new mode 100644 diff --git a/md5sum.txt b/md5sum.txt index 1846c981..00847e92 100644 --- a/md5sum.txt +++ b/md5sum.txt @@ -9,12 +9,12 @@ "episodes/2-keras.Rmd" "9e35ec651717f7323c01c1f4625bace1" "site/built/2-keras.md" "2024-01-23" "episodes/3-monitor-the-model.Rmd" "65a1408b6774e38b951aaa50630ba08a" "site/built/3-monitor-the-model.md" "2024-01-23" "episodes/4-advanced-layer-types.Rmd" "058933200fc97dee980a0b8e80f9c25b" "site/built/4-advanced-layer-types.md" "2024-01-23" -"episodes/5-outlook.Rmd" "fc597b012a3435c9766006b2652e8db0" "site/built/5-outlook.md" "2024-01-23" +"episodes/5-outlook.Rmd" "7782d12cd60bb8fc986f58dcfa44724a" "site/built/5-outlook.md" "2024-01-23" "instructors/bonus-material.md" "d5b6aaee56986ab74e33bb95894cdc0e" "site/built/bonus-material.md" "2024-01-23" "instructors/design.md" "6c13db77f9d69a294398a77da7e9883f" "site/built/design.md" "2024-01-23" "instructors/instructor-notes.md" "b516f8e213b07224e85073bfe47ed3aa" "site/built/instructor-notes.md" "2024-01-23" "instructors/survey-templates.md" "ea5d46e7b54d335f79e57a7bc31d1c5c" "site/built/survey-templates.md" "2024-01-23" -"learners/reference.md" "6e80c34d920c23fd294a69ff5f69f31d" "site/built/reference.md" "2024-01-23" +"learners/reference.md" "ae95aeca6d28f5f0f994d053dc10d67c" "site/built/reference.md" "2024-01-23" "learners/setup.md" "53746145baf2b44786a48b001aeca69f" "site/built/setup.md" "2024-01-23" "profiles/learner-profiles.md" "698c27136a1a320b0c04303403859bdc" "site/built/learner-profiles.md" "2024-01-23" "renv/profiles/lesson-requirements/renv.lock" "2ad3064a33ab4898010b481abbf0ffdb" "site/built/renv.lock" "2024-01-23" diff --git a/reference.md b/reference.md index c4bd4646..0b1d60aa 100644 --- a/reference.md +++ b/reference.md @@ -27,6 +27,7 @@ Here is a (non exhaustive) list of external resources for further study after th - [Unbalanced data](https://towardsdatascience.com/handling-imbalanced-datasets-in-deep-learning-f48407a0e758) - [Unbalanced data in Keras](https://www.tensorflow.org/tutorials/structured_data/imbalanced_data) - [Tensorflow Playground, for visualizing neural networks](http://playground.tensorflow.org/) +- [ChatGPT prompt engineering course](https://learn.deeplearning.ai/chatgpt-prompt-eng/lesson/1/lesson_1) ### Some ML challenges or benchmarks - https://mlcontests.com/