Baconian: a Unified Open-Source Framework for Model-Based

Total Page:16

File Type:pdf, Size:1020Kb

Baconian: a Unified Open-Source Framework for Model-Based Baconian: A Unified Open-source Framework for Model-Based Reinforcement Learning Demonstration Linsen Dong Guanyu Gao Xinyi Zhang School of Computer Science and School of Computer Science and School of Computer Science and Engineering, Nanyang Technological Engineering, Nanyang Technological Engineering, Nanyang Technological University University University Singapore Singapore Singapore [email protected] [email protected] [email protected] Liangyu Chen Yonggang Wen School of Electrical and Electronic School of Computer Science and Engineering, Nanyang Technological Engineering, Nanyang Technological University University Singapore Singapore [email protected] [email protected] ABSTRACT Dyna, GPS, ME, MPC, iLQR, ME- Algorithm Model-Based Reinforcement Learning (MBRL) is one category of PPO, DQN, DDPG, PPO OpenAI Gym, PyBullet, DeepMind Reinforcement Learning (RL) algorithms which can improve sam- Environment pling efficiency by modeling and approximating system dynamics. Control Suite It has been widely adopted in the research of robotics, autonomous Built-in Logging and Visualization, driving, etc. Despite its popularity, there still lacks some sophis- Utility TensorFlow Integration, ticated and reusable open-source frameworks to facilitate MBRL Parameter Management research and experiments. To fill this gap, we develop a flexible User Guide and API References, Open-source MIT License, and modularized framework, Baconian, which allows researchers Support to easily implement a MBRL testbed by customizing or building Benchmark Results Released upon our provided modules and algorithms. Our framework can free users from re-implementing popular MBRL algorithms from Figure 1: Feature list of Baconian. scratch thus greatly save users’ efforts on MBRL experiments. KEYWORDS Existing model-based frameworks are few and have some short- Reinforcement Learning, Model-based Reinforcement Learning, comings. The work in [18] gives a comprehensive benchmark over Open-source Library; state-of-the-art MBRL algorithms, but the implementations are scat- tered across different codebases without a unified implementation, posing obstacles to conduct experiments with it. The work in [5] 1 INTRODUCTION provides the implementations for Guided Policy Search(GPS) [10], Model-Based Reinforcement Learning (MBRL) is proposed to reduce which supports robotics controlling tasks. But it lacks support for sample complexity introduced by model-free Deep Reinforcement other MBRL algorithms. Thus, a unified MBRL open-source frame- arXiv:1904.10762v4 [cs.LG] 16 Mar 2021 Learning (DRL) algorithms [12]. Specifically, MBRL approximates work is in need. To fill this gap, we design and implement aunified the system dynamics with a parameterized model, which can be MBRL framework, Baconian, by trading off the diversity of included utilized for policy optimizing when the training data is very limited MBRL algorithms against the complexity of the framework. Users or costly to obtain in the real world. can reproduce benchmark results or prototype their idea easily Implementing a RL experiments from scratch can be tedious and with it by a minimal amount of codes without understanding the bug-introducing. Luckily, many open-source frameworks have been detailed implementations. Moreover, the design of Baconian not developed to facilitate DRL research, including baselines [3], rllab only benefits the research of MBRL, but is also applicable toother [4], Coach [2], and Horizon [6]. However, these frameworks are types of RL algorithms including model-free algorithms. The code- mainly implemented for model-free DRL methods, and lack enough base is available at https://github.com/cap-ntu/baconian-project. support for MBRL. The demo video is available at https://youtu.be/J6xI6qI3CvE. Proc. of the 19th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2020), B. An, N. Yorke-Smith, A. El Fallah Seghrouchni, G. Sukthankar (eds.) 2 MAIN FEATURES © 2020 International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved. Baconian supports many RL algorithms, test environments, and https://doi.org/doi experiment utilities. We summarize the main features in Fig. 1. Linsen Dong, Guanyu Gao, Xinyi Zhang, Liangyu Chen, and Yonggang Wen Experiment Manager Training Engine Agent Status Policy Algorithm Collector Status Train manage Utilize Agent Experiment Control Dynamic Exploration Setup Settings Launch Flow Strategy User Model flow Experiment Control Recorder Environment Record log Sample Flow Create Environment User Monitor Experiment Logger Plotter Send log Data Run Log file/ Console/ Tensorflow model Figure 3: The procedure to create an MBRL experiment in Figure 2: The system design of Baconian. The system is di- Baconian. Each module is replaceable and configurable to vided into three main modularized components to minimize reduce the effort of building from scratch. the coupling for flexibility and maintainability. 3.1 Experiment Manager The Experiment Manager consists of Experiments Settings, Status State-of-the-Art RL Algorithms. We implement many widely Collector, and Experiment Recorder. Experiments Settings manages used RL algorithms. For model-based algorithms, we implement the creating and initialization of each module. Status Collector con- Dyna[15], ME-TRPO (Model-ensemble Trust Region Policy Opti- trols the status information collected across different modules to mization) [9], MPC (Model Predictive Control)[13], iLQR (Iterative compose a globally shared status that can be used including learn- Linear Quadratic Regulator)[17], etc. Since many model-based al- ing rate decay, exploration strategy scheduling, etc. Experiment gorithms are built upon model-free algorithms, we also implement Recorder will record the information generated from the experi- some popular model-free algorithms including DQN[7], DDPG[11], ment, such as loss, rewards. Such information will be handed to the and PPO[14] in Baconian. Monitoring layer for rendering or saving. Supported Test Environments. To evaluate the performance of RL algorithms, it is a must to support a wide range of test en- vironments. Baconian support OpenAI Gym[1], RoboSchool[8], 3.2 Training Engine DeepMind Control Suite[16]. These test environments cover most Training Engine handles the training process of the MBRL algo- essential tasks in RL community. rithms. The novelty of the design lies in abstracting and encap- Experiment Utilities. Baconian provides many utilities to re- sulating the training process as a Control Flow module, which duce users’ efforts on experiment set-up, hyper-parameter tuning, controls the execution processes of the experiment based on the logging, result visualization, and algorithms diagnosis. We provide user’s specifications, including the agent’s sampling from environ- integration of TensorFlow to support neural network building, train- ment, policy model and dynamics model optimization, and testing. ing, and managing. As the hyper-parameters play a critical role in MBRL algorithms can be complicated[12, 15]. Such abstractions RL experiments, we provide user-friendly parameter management can decouple the tangled and complicated MBRL training processes utility to remove the tedious work of setting, loading, and saving into some independent tasks, which are further encapsulated as the these hyper-parameters. sub-modules of Control Flow module for providing flexibility. Open-source Support. Baconian provide detailed user guides and API references1, so users can hand on Baconian easily and 3.3 Monitor conduct novel MBRL research upon it. We also release some pre- Monitor is responsible for monitoring and recording of the exper- liminary benchmark results in the codebase. iment as it proceeds. This includes recording necessary loggings, printing information/warning/error, and rendering the results. 3 DESIGN AND IMPLEMENTATION Baconian consists of three major components, namely, Experiment 4 USAGE Manager, Training Engine, and Monitor. The system overview of This section presents the procedures to create a MBRL experiment Baconian is shown in Fig. 2. Various design patterns are applied to with the essential modules in Baconian. The procedures are shown decouple the complicated dependencies across different modules to in Fig. 32. For high flexibility, most of modules are customizable. enable the easily extension and programming over the framework. Meanwhile, user can directly adopt built-in benchmark module or codes if customization is unnecessary. 2For more details of how to configure these modules, please see the documentation 1 The documentation can be found at https://baconian-public.readthedocs.io/en/latest/ page https://baconian-public.readthedocs.io/en/latest/step_by_step.html due to the API.html. page limit. Baconian: A Unified Open-source Framework for Model-Based Reinforcement Learning First, the user should create a environment and a RL algorithm [4] Yan Duan, Xi Chen, Rein Houthooft, John Schulman, and Pieter Abbeel. 2016. module with necessary hyper-parameters configured, e.g., neural Benchmarking deep reinforcement learning for continuous control. In Interna- tional Conference on Machine Learning. 1329–1338. network size, learning rate. Algorithm module is usually composed [5] C. Finn, M. Zhang, J. Fu, X. Tan, Z. McCarthy, E. Scharff, and S. Levine. 2016. of a policy module and a dynamics model module depending on Guided Policy Search Code Implementation. (2016). http://rll.berkeley.edu/gps Software available from
Recommended publications
  • Deepfakes & Disinformation
    DEEPFAKES & DISINFORMATION DEEPFAKES & DISINFORMATION Agnieszka M. Walorska ANALYSISANALYSE 2 DEEPFAKES & DISINFORMATION IMPRINT Publisher Friedrich Naumann Foundation for Freedom Karl-Marx-Straße 2 14482 Potsdam Germany /freiheit.org /FriedrichNaumannStiftungFreiheit /FNFreiheit Author Agnieszka M. Walorska Editors International Department Global Themes Unit Friedrich Naumann Foundation for Freedom Concept and layout TroNa GmbH Contact Phone: +49 (0)30 2201 2634 Fax: +49 (0)30 6908 8102 Email: [email protected] As of May 2020 Photo Credits Photomontages © Unsplash.de, © freepik.de, P. 30 © AdobeStock Screenshots P. 16 © https://youtu.be/mSaIrz8lM1U P. 18 © deepnude.to / Agnieszka M. Walorska P. 19 © thispersondoesnotexist.com P. 19 © linkedin.com P. 19 © talktotransformer.com P. 25 © gltr.io P. 26 © twitter.com All other photos © Friedrich Naumann Foundation for Freedom (Germany) P. 31 © Agnieszka M. Walorska Notes on using this publication This publication is an information service of the Friedrich Naumann Foundation for Freedom. The publication is available free of charge and not for sale. It may not be used by parties or election workers during the purpose of election campaigning (Bundestags-, regional and local elections and elections to the European Parliament). Licence Creative Commons (CC BY-NC-ND 4.0) https://creativecommons.org/licenses/by-nc-nd/4.0 DEEPFAKES & DISINFORMATION DEEPFAKES & DISINFORMATION 3 4 DEEPFAKES & DISINFORMATION CONTENTS Table of contents EXECUTIVE SUMMARY 6 GLOSSARY 8 1.0 STATE OF DEVELOPMENT ARTIFICIAL
    [Show full text]
  • Game Playing with Deep Q-Learning Using Openai Gym
    Game Playing with Deep Q-Learning using OpenAI Gym Robert Chuchro Deepak Gupta [email protected] [email protected] Abstract sociated with performing a particular action in a given state. This information is fundamental to any reinforcement learn- Historically, designing game players requires domain- ing problem. specific knowledge of the particular game to be integrated The input to our model will be a sequence of pixel im- into the model for the game playing program. This leads ages as arrays (Width x Height x 3) generated by a particular to a program that can only learn to play a single particu- OpenAI Gym environment. We then use a Deep Q-Network lar game successfully. In this project, we explore the use of to output a action from the action space of the game. The general AI techniques, namely reinforcement learning and output that the model will learn is an action from the envi- neural networks, in order to architect a model which can ronments action space in order to maximize future reward be trained on more than one game. Our goal is to progress from a given state. In this paper, we explore using a neural from a simple convolutional neural network with a couple network with multiple convolutional layers as our model. fully connected layers, to experimenting with more complex additions, such as deeper layers or recurrent neural net- 2. Related Work works. The best known success story of classical reinforcement learning is TD-gammon, a backgammon playing program 1. Introduction which learned entirely by reinforcement learning [6]. TD- gammon used a model-free reinforcement learning algo- Game playing has recently emerged as a popular play- rithm similar to Q-learning.
    [Show full text]
  • Reinforcement Learning with Tensorflow&Openai
    Lecture 1: Introduction Reinforcement Learning with TensorFlow&OpenAI Gym Sung Kim <[email protected]> http://angelpawstherapy.org/positive-reinforcement-dog-training.html Nature of Learning • We learn from past experiences. - When an infant plays, waves its arms, or looks about, it has no explicit teacher - But it does have direct interaction to its environment. • Years of positive compliments as well as negative criticism have all helped shape who we are today. • Reinforcement learning: computational approach to learning from interaction. Richard Sutton and Andrew Barto, Reinforcement Learning: An Introduction Nishant Shukla , Machine Learning with TensorFlow Reinforcement Learning https://www.cs.utexas.edu/~eladlieb/RLRG.html Machine Learning, Tom Mitchell, 1997 Atari Breakout Game (2013, 2015) Atari Games Nature : Human-level control through deep reinforcement learning Human-level control through deep reinforcement learning, Nature http://www.nature.com/nature/journal/v518/n7540/full/nature14236.html Figure courtesy of Mnih et al. "Human-level control through deep reinforcement learning”, Nature 26 Feb. 2015 https://deepmind.com/blog/deep-reinforcement-learning/ https://deepmind.com/applied/deepmind-for-google/ Reinforcement Learning Applications • Robotics: torque at joints • Business operations - Inventory management: how much to purchase of inventory, spare parts - Resource allocation: e.g. in call center, who to service first • Finance: Investment decisions, portfolio design • E-commerce/media - What content to present to users (using click-through / visit time as reward) - What ads to present to users (avoiding ad fatigue) Audience • Want to understand basic reinforcement learning (RL) • No/weak math/computer science background - Q = r + Q • Want to use RL as black-box with basic understanding • Want to use TensorFlow and Python (optional labs) Schedule 1.
    [Show full text]
  • Latest Snapshot.” to Modify an Algo So It Does Produce Multiple Snapshots, find the Following Line (Which Is Present in All of the Algorithms)
    Spinning Up Documentation Release Joshua Achiam Feb 07, 2020 User Documentation 1 Introduction 3 1.1 What This Is...............................................3 1.2 Why We Built This............................................4 1.3 How This Serves Our Mission......................................4 1.4 Code Design Philosophy.........................................5 1.5 Long-Term Support and Support History................................5 2 Installation 7 2.1 Installing Python.............................................8 2.2 Installing OpenMPI...........................................8 2.3 Installing Spinning Up..........................................8 2.4 Check Your Install............................................9 2.5 Installing MuJoCo (Optional)......................................9 3 Algorithms 11 3.1 What’s Included............................................. 11 3.2 Why These Algorithms?......................................... 12 3.3 Code Format............................................... 12 4 Running Experiments 15 4.1 Launching from the Command Line................................... 16 4.2 Launching from Scripts......................................... 20 5 Experiment Outputs 23 5.1 Algorithm Outputs............................................ 24 5.2 Save Directory Location......................................... 26 5.3 Loading and Running Trained Policies................................. 26 6 Plotting Results 29 7 Part 1: Key Concepts in RL 31 7.1 What Can RL Do?............................................ 31 7.2
    [Show full text]
  • ELF Opengo: an Analysis and Open Reimplementation of Alphazero
    ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero Yuandong Tian 1 Jerry Ma * 1 Qucheng Gong * 1 Shubho Sengupta * 1 Zhuoyuan Chen 1 James Pinkerton 1 C. Lawrence Zitnick 1 Abstract However, these advances in playing ability come at signifi- The AlphaGo, AlphaGo Zero, and AlphaZero cant computational expense. A single training run requires series of algorithms are remarkable demonstra- millions of selfplay games and days of training on thousands tions of deep reinforcement learning’s capabili- of TPUs, which is an unattainable level of compute for the ties, achieving superhuman performance in the majority of the research community. When combined with complex game of Go with progressively increas- the unavailability of code and models, the result is that the ing autonomy. However, many obstacles remain approach is very difficult, if not impossible, to reproduce, in the understanding of and usability of these study, improve upon, and extend. promising approaches by the research commu- In this paper, we propose ELF OpenGo, an open-source nity. Toward elucidating unresolved mysteries reimplementation of the AlphaZero (Silver et al., 2018) and facilitating future research, we propose ELF algorithm for the game of Go. We then apply ELF OpenGo OpenGo, an open-source reimplementation of the toward the following three additional contributions. AlphaZero algorithm. ELF OpenGo is the first open-source Go AI to convincingly demonstrate First, we train a superhuman model for ELF OpenGo. Af- superhuman performance with a perfect (20:0) ter running our AlphaZero-style training software on 2,000 record against global top professionals. We ap- GPUs for 9 days, our 20-block model has achieved super- ply ELF OpenGo to conduct extensive ablation human performance that is arguably comparable to the 20- studies, and to identify and analyze numerous in- block models described in Silver et al.(2017) and Silver teresting phenomena in both the model training et al.(2018).
    [Show full text]
  • Applying Deep Double Q-Learning and Monte Carlo Tree Search to Playing Go
    CS221 FINAL PAPER 1 Applying Deep Double Q-Learning and Monte Carlo Tree Search to Playing Go Booher, Jonathan [email protected] De Alba, Enrique [email protected] Kannan, Nithin [email protected] I. INTRODUCTION the current position. By sampling states from the self play OR our project we replicate many of the methods games along with their respective rewards, the researchers F used in AlphaGo Zero to make an optimal Go player; were able to train a binary classifier to predict the outcome of a however, we modified the learning paradigm to a version of game with a certain confidence. Then based on the confidence Deep Q-Learning which we believe would result in better measures, the optimal move was taken. generalization of the network to novel positions. We depart from this method of training and use Deep The modification of Deep Q-Learning that we use is Double Q-Learning instead. We use the same concept of called Deep Double Q-Learning and will be described later. sampling states and their rewards from the games of self play, The evaluation metric for the success of our agent is the but instead of feeding this information to a binary classifier, percentage of games that are won against our Oracle, a we feed the information to a modified Q-Learning formula, Go-playing bot available in the OpenAI Gym. Since we are which we present shortly. implementing a version of reinforcement learning, there is no data that we will need other than the simulator. By training III. CHALLENGES on the games that are generated from self-play, our player The main challenge we faced was the computational com- will output a policy that is learned at the end of training by plexity of the game of Go.
    [Show full text]
  • AI in Focus - Fundamental Artificial Intelligence and Video Games
    AI in Focus - Fundamental Artificial Intelligence and Video Games April 5, 2019 By Isi Caulder and Lawrence Yu Patent filings for fundamental artificial intelligence (AI) technologies continue to rise. Led by a number of high profile technology companies, including IBM, Google, Amazon, Microsoft, Samsung, and AT&T, patent applications directed to fundamental AI technologies, such as machine learning, neural networks, natural language processing, speech processing, expert systems, robotic and machine vision, are being filed and issued in ever-increasing numbers.[1] In turn, these fundamental AI technologies are being applied to address problems in industries such as healthcare, manufacturing, and transportation. A somewhat unexpected source of fundamental AI technology development has been occurring in the field of video games. Traditional board games have long been a subject of study for AI research. In the 1990’s, IBM created an AI for playing chess, Deep Blue, which was able to defeat top-caliber human players using brute force algorithms.[2] More recently, machine learning algorithms have been developed for more complex board games, which include a larger breadth of possible moves. For example, DeepMind (since acquired by Google), recently developed the first AI capable of defeating professional Go players, AlphaGo.[3] Video games have recently garnered the interest of researchers, due to their closer similarity to the “messiness” and “continuousness” of the real world. In contrast to board games, video games typically include a greater
    [Show full text]
  • Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims
    Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims∗ Miles Brundage1†, Shahar Avin3,2†, Jasmine Wang4,29†‡, Haydn Belfield3,2†, Gretchen Krueger1†, Gillian Hadfield1,5,30, Heidy Khlaaf6, Jingying Yang7, Helen Toner8, Ruth Fong9, Tegan Maharaj4,28, Pang Wei Koh10, Sara Hooker11, Jade Leung12, Andrew Trask9, Emma Bluemke9, Jonathan Lebensold4,29, Cullen O’Keefe1, Mark Koren13, Théo Ryffel14, JB Rubinovitz15, Tamay Besiroglu16, Federica Carugati17, Jack Clark1, Peter Eckersley7, Sarah de Haas18, Maritza Johnson18, Ben Laurie18, Alex Ingerman18, Igor Krawczuk19, Amanda Askell1, Rosario Cammarota20, Andrew Lohn21, David Krueger4,27, Charlotte Stix22, Peter Henderson10, Logan Graham9, Carina Prunkl12, Bianca Martin1, Elizabeth Seger16, Noa Zilberman9, Seán Ó hÉigeartaigh2,3, Frens Kroeger23, Girish Sastry1, Rebecca Kagan8, Adrian Weller16,24, Brian Tse12,7, Elizabeth Barnes1, Allan Dafoe12,9, Paul Scharre25, Ariel Herbert-Voss1, Martijn Rasser25, Shagun Sodhani4,27, Carrick Flynn8, Thomas Krendl Gilbert26, Lisa Dyer7, Saif Khan8, Yoshua Bengio4,27, Markus Anderljung12 1OpenAI, 2Leverhulme Centre for the Future of Intelligence, 3Centre for the Study of Existential Risk, 4Mila, 5University of Toronto, 6Adelard, 7Partnership on AI, 8Center for Security and Emerging Technology, 9University of Oxford, 10Stanford University, 11Google Brain, 12Future of Humanity Institute, 13Stanford Centre for AI Safety, 14École Normale Supérieure (Paris), 15Remedy.AI, 16University of Cambridge, 17Center for Advanced Study in the Behavioral Sciences,18Google Research, 19École Polytechnique Fédérale de Lausanne, 20Intel, 21RAND Corporation, 22Eindhoven University of Technology, 23Coventry University, 24Alan Turing Institute, 25Center for a New American Security, 26University of California, Berkeley, 27University of Montreal, 28Montreal Polytechnic, 29McGill University, 30Schwartz Reisman Institute for Technology and Society arXiv:2004.07213v2 [cs.CY] 20 Apr 2020 April 2020 ∗Listed authors are those who contributed substantive ideas and/or work to this report.
    [Show full text]
  • Towards Russian Text Generation Problem Using Openai's GPT-2
    Towards Russian Text Generation Problem Using OpenAI’s GPT-2 Oleksii Shatalov, Nataliya Ryabova National University of Radio Electronics, Nauky av., 14, Kharkiv, 61000, Ukraine Abstract This work is devoted to Natural Language Generation (NLG) problem. The modern approaches in this area based on deep neural networks are considered. The most famous and promising deep neural network architectures that are related to this problem are considered, in particular, the most popular free software solutions for NLG based on Transformers architecture with pre-trained deep neural network models GPT-2 and BERT. The main problem is that the main part of already existing solutions is devoted to the English language. But there are few models that are able to generate text in Russian. Moreover, the text they generate often belongs to a general topic and not about a specific subject area. The object of the study is the generation of a contextually coherent narrow-profile text in Russian. Within the framework of the study, a model was trained for generating coherent articles of a given subject area in Russian, as well as a software application for interacting with it. Keywords 1 Natural Language Generation, Natural Language Processing, Transformers Architecture, Deep Learning, Transfer Learning, GPT-2 1. Introduction The current rate of growth of content is so great that organizations are beginning to fail to keep up with their own set of speeds. Editors and copywriters do not have time to create new texts from scratch, think over ideas for new publications so that they are original. Hiring a large staff of additional staff can significantly increase the costs of the company, which will lead to lower profits.
    [Show full text]
  • The Future Just Arrived
    BLOG The Future Just Arrived Brad Neuman, CFA Senior Vice President Director of Market Strategy What if a computer were actually smart? You wouldn’t purpose natural language processing model. The have to be an expert in a particular application to interface is “text-in, text-out” that goal. but while the interact with it – you could just talk to it. You wouldn’t input is ordinary English language instructions, the need an advanced degree to train it to do something output text can be anything from prose to computer specific - it would just know how to complete a broad code to poetry. In fact, with simple commands, the range of tasks. program can create an entire webpage, generate a household income statement and balance sheet from a For example, how would you create the following digital description of your financial activities, create a cooking image or “button” in a traditional computer program? recipe, translate legalese into plain English, or even write an essay on the evolution of the economy that this strategist fears could put him out of a job! Indeed, if you talk with the program, you may even believe it is human. But can GPT-3 pass the Turing test, which assesses a machine’s ability to exhibit human-level intelligence. The answer: GPT-3 puts us a lot closer to that goal. I have no idea because I can’t code. But a new Data is the Fuel computer program (called GPT-3) is so smart that all you have to do is ask it in plain English to generate “a The technology behind this mind-blowing program has button that looks like a watermelon.” The computer then its roots in a neural network, deep learning architecture i generates the following code: introduced by Google in 2017.
    [Show full text]
  • (GPT-3) in Healthcare Delivery ✉ Diane M
    www.nature.com/npjdigitalmed COMMENT OPEN Considering the possibilities and pitfalls of Generative Pre- trained Transformer 3 (GPT-3) in healthcare delivery ✉ Diane M. Korngiebel 1 and Sean D. Mooney2 Natural language computer applications are becoming increasingly sophisticated and, with the recent release of Generative Pre- trained Transformer 3, they could be deployed in healthcare-related contexts that have historically comprised human-to-human interaction. However, for GPT-3 and similar applications to be considered for use in health-related contexts, possibilities and pitfalls need thoughtful exploration. In this article, we briefly introduce some opportunities and cautions that would accompany advanced Natural Language Processing applications deployed in eHealth. npj Digital Medicine (2021) 4:93 ; https://doi.org/10.1038/s41746-021-00464-x A seemingly sophisticated artificial intelligence, OpenAI’s Gen- language texts. Although its developers at OpenAI think it performs erative Pre-trained Transformer 3, or GPT-3, developed using well on translation, question answering, and cloze tasks (e.g., a fill-in- computer-based processing of huge amounts of publicly available the-blank test to demonstrate comprehension of text by providing textual data (natural language)1, may be coming to a healthcare the missing words in a sentence)1, it does not always predict a 1234567890():,; clinic (or eHealth application) near you. This may sound fantastical, correct string of words that are believable as a conversation. And but not too long ago so did a powerful computer so tiny it could once it has started a wrong prediction (ranging from a semantic fit in the palm of your hand.
    [Show full text]
  • Patent Claim Generation by Fine-Tuning Openai GPT-2
    Patent Claim Generation by Fine-Tuning OpenAI GPT-2 Jieh-Sheng Lee and Jieh Hsiang Department of Computer Science and Information Engineering National Taiwan University {d04922013, jhsiang}@ntu.edu.tw Abstract (Bidirectional Encoder Representations from Transformers) [4] has become the best practice In this work, we focus on fine-tuning an for state-of-the-art results. GPT-2 is the successor OpenAI GPT-2 pre-trained model for to GPT. Although both GPT-2 and BERT are generating patent claims. GPT-2 has capable of text generation, Wang and Cho [5] demonstrated impressive efficacy of pre- trained language models on various tasks, found that GPT-2 generations are of better particularly coherent text generation. quality. In fact, GPT-2 is claimed to be so Patent claim language itself has rarely powerful that the risk of its malicious use is high. been explored in the past and poses a For this reason, OpenAI decided to keep its unique challenge. We are motivated to largest model (1.5B parameters) closed so that generate coherent patent claims there is more time to discuss its ramifications. automatically so that augmented inventing In this work, we generated patent claims by might be viable someday. In our fine-tuning the released 345M medium version implementation, we identified a unique [6]. Overall we are impressed by how coherent language structure in patent claims and and complicate the generated patent claims could leveraged its implicit human annotations. We investigated the fine-tuning process by be, although not all text are generated equally in probing the first 100 steps and observing terms of quality.
    [Show full text]