-
Constrained-Text-Generation-Studio
Code repo for "Most Language Models can be Poets too: An AI Writing Assistant and Constrained Text Generation Studio" at the (CAI2) workshop, jointly held at (COLING 2022)
-
InfluxDB
Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.
-
I did my own experiment with https://chat.openai.com/ recently.
I asked it to tell me about myself, based on my GitHub profile. Its response was detailed, well written, and wrong. It told me that I had developed several tools that I could very plausibly have developed -- but I didn't. In particular, it told me that I had written something called "wgrep", a version of grep for Windows that works with Windows file formats and binary files. That's just the kind of thing I might have done, but it doesn't exist. (GNU grep works well on Windows.)
When I asked it when I had worked at one of my previous employers, it said it consulted by Linkedin profile, but it got the dates complete wrong. It said that I had worked on several projects -- all of which are things that interest me, but none of which I actually worked on.
If a human came up with this, I'd say they were lying, but ChatGPT doesn't have the awareness necessary to lie. The closest analogy I can think of is a reckless disregard for the truth.
https://github.com/hellisotherpeople/constrained-text-genera...
Just ban the damn tokens and try again. I wish that folks had more intuition around tokenization, and why LLMs struggle to follow syntactic, lexical, or phonetic constraints.
https://github.com/hellisotherpeople/constrained-text-genera...
Just ban the damn tokens and try again. I wish that folks had more intuition around tokenization, and why LLMs struggle to follow syntactic, lexical, or phonetic constraints.
Did you arrive at this certainty through reading something other than what OpenAI has published? The document [0] that describes the training data for GPT-2 makes this assertion hilarious to me.
[0]: https://github.com/openai/gpt-2/blob/master/model_card.md#da...