I’ve recently found success by making YouTube videos documenting my process and problem solving methods. Much of my work comes from when I was consulting for startups. I’m no longer interested in consulting work, however I still need interesting problems to solve.
I think I’ve found a compromise. If you send me your research problem or business problem, I’ll solve it for you for free, the data and code will be published under the MIT license. The exchange is that you get free labor and insights while I get interesting research problems and ideas for content.
Important note: I have not violated any NDAs. Most of the content ideas I’ve made videos about came from a single phone call with one startup or another. They just wanted to chat, and they did not share any proprietary data or code with me. I do have a few NDAs that expire next year that will make for some interesting content.
When you say “detecting” what do you mean? just identifying segments of text? Or do you want to fix it?
For the second one - detecting incorrect information - that will be much harder as GPT-3 doesn’t necessarily have tight integration with a source of ground truth. However, I suspect that it can handle some situations.
It would be helpful to have full sentences. Was this data notated by hand? Do you already have a pipeline that isolates dates? It seems to me you might be approaching this in too oldschool of a fashion. GPT-3 can extract dates from unstructured text. The problem data you have here is pretty trivial, to be honest
I don’t need this issue solved currently, but I occasionally do freelance work, and earlier in the year I had a fairly confusing contract to sign. I’m not good at deciphering legalese and used GPT-3 to help me with that.
I basically shortened the various points made in the contract into more easily understandable sentences. I’m extremely new to using tools like GPT-3, I did get it to work… somewhat, but thought it would make life easier if someone else could make an online tool where someone could copy/paste their hard-to-understand legalese of a contract, and that tool would decipher it for them.
An additional thought was if Open AI (or someone) created an embeddable token purchase app that could be embedded into the webpage with the legalese tool (or any GPT-3 tool), something like how paypal has an embedable app that allows people to purchase goods from a personal website, but instead you would purchase tokens to use the app (and just the amount needed!), not a preset amount.
I’m guessing that in doing this, the creator of that token app could take a cut of the sale.
Being unemployed right now, I’d love to do this myself, but I have no idea how to go about this, and it sounds like you do.
(I wouldn’t feel comfortable sharing the contracts that I had to sign, but it sounds like you probably have access to some.)
Eight years ago, one of my dreams that I plan to just that, using artificial intelligence that can be a lot of problems to solve such a problem, filmmakers, animators and architects and engineers and designers and the idea for a fantasy-hearted woman I Unfortunately, because expertise in the field of artificial intelligence not many places I contacted any answer convincingly Feb nothing I saw was that I could say a program I correct when saying audio to multiple threads created, for example, if I want a story to write epic and learn the story of the action without the need for a director and cameraman, and various elements of this program is that by telling me all the stereotypes that I have created and the story of creation. this is one of my ideas was to hopefully Google so that the international community can do to use it
I have many and various new ideas. Eight years ago, one of my dreams that I plan to just that, using artificial intelligence that can be a lot of problems to solve such a problem, filmmakers, animators and architects and engineers and designers and the idea for a fantasy in the mind I Unfortunately, because expertise in the field of artificial intelligence not many places I contacted any answer convincing me not my idea was that I could tell my own and I correct when saying the audio to a variety of topics to create, for example, if I want to story to write epic and learn the story of the action without the need for a director and cameraman, and various elements of this program could by telling me all the stereotypes Vpara meters and actors that I have created and the story of creation. this is one of my ideas Google could do was hope the international community to use it
We’re trying to categorize company descriptions (and in a separate model) the company’s homepage URL text (text on the homepage) into categories so we can track trends and find similarities
Primarily trend analysis among startups. As new startups launch, we get information about it from it’s homepage as well as other generic data such as description from linkedin. We’re trying to do topic modelling to better see QoQ, YoY trends and give user ability to sift through companies based on a topic
We currently use semantic search to let users search through companies but surfacing relevant topics may inspire users to find new trends and categories they wouldn’t have naturally thought of
