james-revisoai

19 ms·

181.
by james-revisoai 4y ago
Really cool to see! I totally agree that GPT models will revolutionise education, and a few thousand students are using a tool I've launched in that space that runs on PDFs and Youtube videos. It's specifically optimised LLMs to a
182.
by james-revisoai 4y ago
Here are the types of images your users might be searching for: Examples of a class of objects - say "Lion" - your best friend is CLIP, trained on captions and naturally weighted towards popular images. Conceptual ideas, abstract
183.
by james-revisoai 4y ago
because of this issue, when you upload a video to a service I run, it populates and biases the speech-to-text (ASR... using whisper 'prompt' parameter) towards terms in the title and descriptions of the video (from youtube) - I ha
184.
by james-revisoai 4y ago
Can you give an example as to what you mean please?
185.
by james-revisoai 4y ago
don't want to jump on your brigade, but at AnyQuestions.ai we specialise in quality PDF and transcript processing for AI-answer purposes (supporting AI answers with citations, just like your tool for documentation). This comes from 3 y
186.
by james-revisoai 4y ago
Just FYI, my last comment has now been slightly outdated by this new paper today: https://arxiv.org/abs/2301.12652
187.
by james-revisoai 4y ago
Yep, you could pull a saved version of the model weights for your specific data and restore it later, which is what a lot of the finetuned models on huggingface do under the hood(and probably what OpenAI does when you finetune on the API).
188.
by james-revisoai 4y ago
Yeah that's it exactly, and a major thing people run into. Finetuning won't add "knowledge", it's changing the bias likelihood going forward for each tokens generation(with context to tokens around it in your traini
189.
by james-revisoai 4y ago
For 1. I think you may find work with semantic text chunking interesting, and the possibility of overlaying embeddings with one another (e.g. powering fragen.co.uk youtube search, the transcriptions whenever you click the "play" i
190.
by james-revisoai 4y ago
Behind this thinking - an economic incentive to access knowledge and answers quicker that is inevitable - we have built a tool that verifies statements using source materials. You have to put the legwork in to upload your PDFs/web page
191.
by james-revisoai 4y ago
Professional website, nice. Can I use this to overcome some scaling/dev ops problems? E.g. I have a list of errors a number of users experience, but are tricky for my team to act on when we have a lot of customer support to do etc. And
192.
by james-revisoai 4y ago
Hey! I'm working on building this vision at fragen.co.uk - which can take in Youtube videos (trascribed by whisper + postprocessed into useful chunks), PDFs (properly OCR'd - understands bullet points etc) and webpages (also OCR&#
193.
by james-revisoai 4y ago
I had this feeling too, so I began feeding text from my documents and videos into the same GPT engine. It results in much more relevant and grounded answers(well... grounded to whatever you documents claim, for better or worse). If you want
194.
by james-revisoai 4y ago
Yes - if the core innovations in ChatGPT hadn't came with the base GPT-3 model, we'd see that too. Just imagine if they had released a model which transformed a page of search engine result snippets into the answers ChatGPT gives
195.
by james-revisoai 4y ago
You still interested in this? I'd be keen to chat to you, worked on a searchable transcript provider for educational youtube videos (likewise, unfortunately pre-whisper, so I did a lot of work with sentence completion perplexity and rp
196.
Show HN: Automatically turn few-word Topics into a Quiz(revision.ai)
3 points by james-revisoai 5y ago|1 comments
197.
by james-revisoai 5y ago
Hi HN! This is a tool for learners-in-a-hurry, built to make it easier to engage with learning. It’s motivated by the Testing Effect, which shows studying by reading is not as good as using quizzes/flashcards (grade difference ~20%).