) Chatgpt

[Tisham Candella]

Wealso just released a TypeScript chatgpt-plugin package which contains helpers and examples to make it as easy as possible to start building your own ChatGPT Plugins in JS/TS. Even if you don't have developer access to ChatGPT Plugins yet, you can still use the chatgpt-plugin repo to get a head start on building your own plugins locally.

Please upgrade to chatgpt@latest (at least v4.0.0). The updated version is significantly more lightweight and robust compared with previous versions. You also don't have to worry about IP issues or rate limiting.

By default, the response is streamed to stdout, the results are stored in a local config file, and every invocation starts a new conversation. You can use -c to continue the previous conversation and --no-stream to disable streaming.

ChatGPTAPI - Uses the gpt-3.5-turbo model with the official OpenAI chat completions API (official, robust approach, but it's not free). You can override the model, completion params, and system message to fully customize your assistant.

ChatGPTUnofficialProxyAPI - Uses an unofficial proxy server to access ChatGPT's backend API in a way that circumvents Cloudflare (uses the real ChatGPT and is pretty lightweight, but relies on a third-party server and is rate-limited)

Note: We strongly recommend using ChatGPTAPI since it uses the officially supported API from OpenAI and it also supports gpt-4. We will likely remove support for ChatGPTUnofficialProxyAPI in a future release.

To use ChatGPTUnofficialProxyAPI, you'll need an OpenAI access token from the ChatGPT webapp. To do this, you can use any of the following methods which take an email and password and return an access token:

Note: using a reverse proxy will expose your access token to a third-party. There shouldn't be any adverse effects possible from this, but please consider the risks before using this method.

Most of the demos use ChatGPTAPI. It should be pretty easy to convert them to use ChatGPTUnofficialProxyAPI if you'd rather use that approach. The only thing that needs to change is how you initialize the api with an accessToken instead of an apiKey.

Note that persisting message is required for remembering the context of previous conversations beyond the scope of the current Node.js process, since by default, we only store messages in memory. Here's an external demo of using a completely custom database solution to persist messages.

Artificial intelligence (AI) language models, such as ChatGPT, may be used for any assignment with appropriate citation. Examples of citing AI language models are available at: libguides.umn.edu/chatgpt [or provide an alternative reference appropriate for your class]. You are responsible for fact checking statements composed by AI language models.

Artificial intelligence (AI) language models, such as ChatGPT, may be used for [assignment types A, B & C] with appropriate citation, but not for [assignment types D, E & F]. If you are in doubt as to whether you are using AI language models appropriately in this course, I encourage you to discuss your situation with me. Examples of citing AI language models are available at: libguides.umn.edu/chatgpt [or provide an alternative reference appropriate for your class]. You are responsible for fact checking statements composed by AI language models.

"Scholastic dishonesty means plagiarism; cheating on assignments or examinations, including the unauthorized use of online learning support and testing platforms; engaging in unauthorized collaboration on academic work, including the posting of student-generated coursework on online learning support and testing platforms not approved for the specific course in question; taking, acquiring, or using course materials without faculty permission, including the posting of faculty-provided course materials on online learning and testing platforms; ..."

Artificial intelligence (AI) language models, such as ChatGPT, and online assignment help tools, such as Chegg, are examples of online learning support platforms: they can not be used for course assignments except as explicitly authorized by the instructor. The following actions are prohibited in this course [remove bullets as necessary]:

Any assignment content composed by any resource other than you, regardless of whether that resource is human or digital, must be attributed to the source through proper citation. (Examples of citing content composed by digital tools are presented in: libguides.umn.edu/chatgpt [or provide an alternative reference appropriate for your class].)

Now because I am not a chatbot but instead a living, breathing human who in theory needs to eat to survive, I should remind you that if you like what you are reading here you can help by sharing what I write (for I rely on word of mouth for my audience) and by supporting me on Patreon. And if you want updates whenever a new post appears, you can click below for email updates or follow me on twitter (@BretDevereaux) for updates as to new posts as well as my occasional ancient history, foreign policy or military history musings, assuming there is still a Twitter by the time this post goes live.

But I think we want to start by discussing what ChatGPT is and what it is not; it is the latter actually that is most important for this discussion. The tricky part is that ChatGPT and chatbots like it are designed to make use of a very influential human cognitive bias that we all have: the tendency to view things which are not people as people or at least as being like people. We all do this; we imagine our pets understand more than they can, have emotions more similar to ours than they do,3 or that inanimate objects are not merely animate but human in their feelings, memories and so on. We even imagine that the waves and winds are like people too and assign them attributes as divine beings with human-like emotions and often human-like appearances. We beg and plead with the impersonal forces of the world like we would with people who might be moved by those emotions.

It is crucial to note, however, what the data is that is being collected and refined in the training system here: it is purely information about how words appear in relation to each other. That is, how often words occur together, how closely, in what relative positions and so on. It is not, as we do, storing definitions or associations between those words and their real world referents, nor is it storing a perfect copy of the training material for future reference. ChatGPT does not sit atop a great library it can peer through at will; it has read every book in the library once and distilled the statistical relationships between the words in that library and then burned the library.

Writing an essay thus involves a number of steps, of which communication is merely the last. Ideally, the essay writer has first observed their subject, then drawn some sort of analytical conclusion about that subject,5 then organized their evidence in a way that expresses the logical connections between various pieces of evidence, before finally communicating that to a reader in a way that is clear and persuasive.

But the core work of the essay? This ChatGPT cannot do. And importantly it is not some capacity which merely requires iterative improvements on the product. While ChatGPT can fake an original essay, the jump from faking that essay to writing an actually original thought certainly looks like it would require a completely different program, one capable of observing the real world, analyzing facts about it and then reaching conclusions.

Thus the last and most important thing I am trying to train is not the form of the essay nor its content, but the basic skills of having a thought and putting it in a box that we outlined earlier. Even if your job or hobbies do not involve formal writing, chances are (especially if your job requires a college degree) you are still expected to observe something real, make conclusions about it and then present those conclusions to someone else (boss, subordinates, co-workers, customers, etc.) in a clear way, supported by convincing evidence if challenged. What we are practicing then is how to have good thoughts, put them in good boxes and then effectively hand that box to someone else. That can be done in a formal written form (the essay), in informal writing (emails, memos, notes, Slack conversations), or verbally (speeches, but also arguments, debates and discussions). The skills of having the idea, supporting it with evidence, organizing that evidence effectively to be understood and then communicating that effectively are transferable and the most important skills that are being practiced when a student writes an essay.

To which some boosters of the technology respond that what I should really be doing is training students on how to most effectively use ChatGPT as a tool. But it is not clear to me that ChatGPT functions well as a tool for any part of this process. One suggestion is to write an outline and then feed that into ChatGPT to generate a paper, but that fails to train the essential communication component of the assignment and in any case, ChatGPT is actually pretty bad at the nuts of and bolts of writing paragraphs. Its tendency in particular to invent facts or invent non-existent sources to cite makes it an enormous liability here; it is a very bad research tool because it is unreliable. Alternately the suggestion is that students could use ChatGPT to produce an essay they edit to fit or an outline they fill in; both problems run into the issue that the student is now trying to offload the most important part of the task for them to learn: the actual thinking and analysis. And the crucial thing to note is that the skill that is not being trained in both cases is a skill that current large language models like ChatGPT cannot perform or perform very poorly.12

3a8082e126