44
votes
Accepted
Why do ChatGPT “jailbreaks” work?
"Jailbreaks" work for a variety of reasons:
A lot of the setup that turns an LLM instance into a polite, well-behaved chatbot is actually just a hidden piece of starting text (a "pre-...
- 33.3k
25
votes
Why do ChatGPT “jailbreaks” work?
Let's step back for a moment and consider your assertion:
It seems like if they are able to develop such impressive AI models
This implies that you are thinking of these models as being programmed ...
- 351
24
votes
Accepted
What exactly are the "parameters" in GPT-3's 175 billion parameters and how are they chosen/generated?
Parameters is a synonym for weights, which is the term most people use for a neural networks parameters (and indeed in my experience it is a term that machine learners will use in general whereas ...
- 5,030
13
votes
Accepted
Why does ChatGPT create fake code?
ChatGPT is a generalised model. It does not understand any code. It does not know it is creating fake functions. It doesn't know that it isn't solving the question asked, because it doesn't know what ...
- 368
13
votes
Accepted
Are GPT-3.5 series models based on GPT-3?
I'll complement nbro's answer with this great visual summary by Yao Fu <[email protected]>:
- 3,267
10
votes
How does ChatGPT respond to novel prompts and commands?
Text continuation has the same reasons to work in any context, be it the middle of a sentence, after a question or after instructions. Following your example, the same word sequence could be a good ...
8
votes
How to assess if OpenAI's ChatGPT chatbot has a human in the loop?
Until you can prove that OpenAI has HITL in ChatGPT, it is just an idea with no basis. It's not up to us to disprove it, it's up to you to prove it.
Let me address your points:
You seem to be basing ...
- 189
7
votes
Are GPT-3.5 series models based on GPT-3?
ChatGPT has not been trained from scratch. ChatGPT is a fine-tuned version of a model from the GPT-3.5 series. OpenAI writes
ChatGPT is fine-tuned from a model in the GPT-3.5 series, which finished ...
- 41.4k
7
votes
Why do ChatGPT “jailbreaks” work?
"Jailbreaks" work because there are no "jails" in the AI model.
The models are enormous collections of interconnected statistics. Your "prompt" starts the model to ...
- 171
6
votes
OpenAI ChatGPT gives a network error at long responds. How can I fix it?
I resolved it by asking it to show me the first 10-20 lines, then the next 10 and so on until the end. In my case those were lines of code so it worked for me.
6
votes
Accepted
Would self-hosting ChatGPT be feasible, w.r.t. computation costs?
Here, I assume the ChatGPT architecture is similar to the published GPT-3 model [1].
The Transformer architecture used for GPT3 is much more memory-bound than compute-bound at a high parameter count ...
- 266
6
votes
Is GPT-4 based on GPT-3 or was it trained from the scratch?
GPT-4 is largely based on GPT-3. From the GPT-4 Technical Report:
GPT-4 is a Transformer-style model [39]
The transformer-style model originates from the paper Attention Is All You Need, which lays ...
- 1,160
6
votes
How do open source LLMs compare to GPT-4?
The remarkable performance of GPT 4 is due to the massive size of its architecture and the amount of data it was trained on, which costs a lot of money. Few organizations have the hardware resources ...
- 1,987
6
votes
Accepted
How do you deal with movement inertia in an environment after a step?
The simplest answer is that the inertia (or velocity, if the mass of the ball is not a variable) should be part of the observable state that the agent has access to. In RL there are usually two parts ...
- 33.3k
5
votes
Accepted
How powerful is OpenAI's Gym and Universe in board games area?
OpenAI's Gym is a standardised API, useful for reinforcement learning, applied to a range of interesting environments many of which you can then access for free with little effort. It is very simple ...
- 33.3k
5
votes
Accepted
Has anyone been able to solve OpenAI's hardcore bipedal walker with their implementation of DDPG?
You may be very interested to know that there was a bug in the v2 Lidar tracing, making the agent think there were phantom objects, and sometimes intersecting with its own legs:
https://github.com/...
- 206
5
votes
Accepted
Should the exploration rate be reset after each trial in Q-learning?
The exploration rate, typically parameterized as epsilon / ε, can be changed on every trial. It depends on the complexity of the model and the goals.
The simplest thing to do is keep exploration rate ...
- 281
5
votes
Creating a support chat bot for my business
First and foremost, do not use GPT/OpenAI for customer-facing applications. You end up with a mess. GPT is great for creative work, but not for production. GPT is a probabilistic language model, and ...
- 53
4
votes
Accepted
OpenAI Gym: How is "experience" stored?
This is sort of a tricky question to answer, by results you could mean 2 possible things. History, which is a culmination of actions, observations and rewards. Or the value func/policy which is the ...
- 1,260
4
votes
InstructGPT: What is the sigma in the loss function and why $\log(\cdot)$ is being used?
According to this guide, the sigma in this formula refers to the sigmoid activation function. The guide does not tell exactly why the sigmoid function is used here, so I will try to give a full ...
- 141
4
votes
Would self-hosting ChatGPT be feasible, w.r.t. computation costs?
Suppose the pre-trained, current date (2023-02-04) ChatGPT model was released open source, would it be feasible for regular users to interact with the model on a self-hosted computer?
No, assuming it'...
- 3,267
4
votes
Accepted
Is GPT-4 based on GPT-3 or was it trained from the scratch?
GPT-4 is a transformer like GPT-3 and any other GPT. The training is certainly new, because it has a different size, you just can not transfer GPT-3 weights into an GPT-4 to continue training.
The ...
- 767
4
votes
Accepted
OpenAI: What is the difference between model "gpt-3.5-turbo" and "gpt-3.5-turbo-0301"?
Taken from here: https://platform.openai.com/docs/models/gpt-3-5
I think its literally an update but the specifics of what that updates are I do not know I looked through the documentation but this ...
- 93
4
votes
How do open source LLMs compare to GPT-4?
How do open source LLMs compare to GPT-4?
https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard has a leaderboard containing both open source LLMs and GPT-4 (and GPT-3.5-turbo):
Model
⭐ ...
- 3,267
4
votes
How do you deal with movement inertia in an environment after a step?
In RL usually a "discretized" time MDP is used in case of continuous time problems, with the idea that if we evaluate enough time the policy, we can approximate it the continuous case. This ...
- 2,632
4
votes
Accepted
What is accelerated years in describing the amount of the training time?
The statement in which you mentioned that "GPT-3 took 405 V100 years to train" refers to the computational resources utilized in training the GPT-3 model. Specifically measured in terms of ...
- 576
3
votes
Can the opponent's turn affect the reward for a DQN agent action?
Does the opponent's turn affect the calculated rewards?
Yes, in general it can. Obvious case, in a two player game where the opponent could win or lose on their turn, but has other options.
As ...
- 33.3k
3
votes
Is a neural network able to optimize itself for speed?
Yes and No.
Yes, it is possible to achieve that result. But instead of a self-optimizing Neural Network, I'd recommend another approach:
1. Don't let training time interfere.
If you are trying to ...
- 854
3
votes
Accepted
OpenAI Gym: Multiple actions in one step
What I was looking for is multi-agent RL, where I have multiple RL agents, each controlling actions of one user. All RL agents/user make an action in each environment step and each get their own ...
- 183
3
votes
How do I use GPT-2 to summarise text?
In short: It depends.
Where will you run it?
On Premises: You may want want to run in your own environment.
IaaS: GPT models are often too big, so people might prefer to setup a different server for ...
- 854
Only top scored, non community-wiki answers of a minimum length are eligible