ChatGPT: Optimizing Language Models for Dialogue

We’ve trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests. ChatGPT is a sibling model to InstructGPT, which is trained to follow an instruction in a prompt and provide a detailed response.

We are excited to introduce ChatGPT to get users’ feedback and learn about its strengths and weaknesses. During the research preview, usage of ChatGPT is free. Try it now at


We trained this model using Reinforcement Learning from Human Feedback (RLHF), using the same methods as InstructGPT, but with slight differences in the data collection setup. We trained an initial model using supervised fine-tuning: human AI trainers provided conversations in which they played both sides—the user and an AI assistant. We gave the trainers access to model-written suggestions to help them compose their responses.

To create a reward model for reinforcement learning, we needed to collect comparison data, which consisted of two or more model responses ranked by quality. To collect this data, we took conversations that AI trainers had with the chatbot. We randomly selected a model-written message, sampled several alternative completions, and had AI trainers rank them. Using these reward models, we can fine-tune the model using Proximal Policy Optimization. We performed several iterations of this process.

ChatGPT is fine-tuned from a model in the GPT-3.5 series, which finished training in early 2022. You can learn more about the 3.5 series here. ChatGPT and GPT 3.5 were trained on an Azure AI supercomputing infrastructure.


  • ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this issue is challenging, as: (1) during RL training, there’s currently no source of truth; (2) training the model to be more cautious causes it to decline questions that it can answer correctly; and (3) supervised training misleads the model because the ideal answer depends on what the model knows, rather than what the human demonstrator knows.
  • ChatGPT is sensitive to tweaks to the input phrasing or attempting the same prompt multiple times. For example, given one phrasing of a question, the model can claim to not know the answer, but given a slight rephrase, can answer correctly.
  • The model is often excessively verbose and overuses certain phrases, such as restating that it’s a language model trained by OpenAI. These issues arise from biases in the training data (trainers prefer longer answers that look more comprehensive) and well-known over-optimization issues.12
  • Ideally, the model would ask clarifying questions when the user provided an ambiguous query. Instead, our current models usually guess what the user intended.
  • While we’ve made efforts to make the model refuse inappropriate requests, it will sometimes respond to harmful instructions or exhibit biased behavior. We’re using the Moderation API to warn or block certain types of unsafe content, but we expect it to have some false negatives and positives for now. We’re eager to collect user feedback to aid our ongoing work to improve this system.

Iterative deployment

Today’s research release of ChatGPT is the latest step in OpenAI’s iterative deployment of increasingly safe and useful AI systems. Many lessons from deployment of earlier models like GPT-3 and Codex have informed the safety mitigations in place for this release, including substantial reductions in harmful and untruthful outputs achieved by the use of reinforcement learning from human feedback (RLHF).

Copy from Source:

Long-Range Transformers for Dynamic Spatiotemporal Forecasting

Multivariate Time Series Forecasting focuses on the prediction of future values based on historical context. State-of-the-art sequence-to-sequence models rely on neural attention between timesteps, which allows for temporal learning but fails to consider distinct spatial relationships between variables. In contrast, methods based on graph neural networks explicitly model variable relationships. However, these methods often rely on predefined graphs and perform separate spatial and temporal updates without establishing direct connections between each variable at every timestep. This paper addresses these problems by translating multivariate forecasting into a spatiotemporal sequence formulation where each Transformer input token represents the value of a single variable at a given time. Long-Range Transformers can then learn interactions between space, time, and value information jointly along this extended sequence. Our method, which we call Spacetimeformer, achieves competitive results on benchmarks from traffic forecasting to electricity demand and weather prediction while learning fully-connected spatiotemporal relationships purely from data.


3D Cellular Automata

This is a project that ressembles the classic Game of Life, but implemented in a 3D world. The project needs OpenGL 4.5 or greater to render, and currently only compiles under Linux (although it would be relatively easy to change CMakeLists to compile for Windows/macOS*(apple discontinued OpenGL support in favor of Metal)).

Cellular Automaton 3D



Flexoskeleton printing: Fabricating flexible exoskeletons for insect-inspired robots

Insects typically have a variety of complex exoskeleton structures, which support them in their movements and everyday activities. Fabricating artificial exoskeletons for insect-inspired robots that match the complexity of these naturally-occurring structures is a key challenge in the field of robotics.
Flexoskeleton printing: fabricating flexible exoskeletons for insect-inspired robots

Although researchers have proposed several  and techniques to produce exoskeletons for insect-inspired robots, many of these methods are extremely complex or rely on expensive equipment and materials. This makes them unfeasible and difficult to apply on a wider scale.

With this in mind, researchers at the University of California in San Diego have recently developed a new process to design and fabricate components for insect-inspired robots with  structures. They introduced this process, called flexoskeleton printing, in a paper prepublished on arXiv.

“Inspired by the insect exoskeleton, we present a new  process called ‘flexoskeleton’ printing that enables rapid and accessible fabrication of hybrid rigid/soft robots,” the researchers wrote in their paper.

So far, hybrid robots with both rigid and soft components have been typically built using expensive materials and 3-D printers, as well as multi-step casting and machine processes. In their study, the research team at UC San Diego set out to create a new fabrication method that is cheaper and easier to use.

Flexoskeleton printing: fabricating flexible exoskeletons for insect-inspired robots

a) A figure explaining how the printing process introduced by the researchers works. b) A four-legged robot created using the researchers’ method, immediately after printing on clear PC layer. c) The four legged robot after release from the PC layer. Credit: Jiang, Zhou & Gravish.

Flexoskeleton printing, the method they developed, relies on an adaptation of a consumer grade fused deposition material (FDM) 3-D printer, which provides an extremely strong bond strength between the deposited material and the printer’s flexible base layer. This process can be used to create exoskeletons for insect-inspired robots with different shapes and morphologies.

Remarkably, the fabrication approach proposed by the researchers can be used by both novice and expert users, as it is fairly straightforward and easy to understand. It is also far more affordable than alternative fabrication methods, as the materials and equipment it relies on are considerably cheap and readily available.

In their study, the team demonstrated the feasibility of their approach by using it to design and test a wide variety of canonical flexoskeleton elements. They then combined all the elements they produced into a walking four-legged  with a flexible exoskeleton structure.

“The approach we have developed relies heavily on the interrelationships between three dimensional geometry of surface features and their contributions to the local mechanical properties of that component,” the researchers wrote in their paper. “We envision that this method will enable a new class of bio-inspired robots with focus on the interrelationships between  and locomotion.”

In the future, the new design and fabrication process devised by this team of researchers could enable the development of numerous insect-inspired robots. As the technique is far more straightforward and affordable than most existing methods, it could also make existing or new robots easier to scale up, increasing their chances of being produced in larger quantities and appearing on the market.


More information: Flexoskeleton printing for versatile insect-inspired robots. arXiv:1911.06897 [cs.RO].