Two apps: virtual writing assistant, and a verified news app

Community
1

Greetings OpenAi community.

I’m a novice to this field (not a coder) but fascinated by AI and its capabilities; I’ve been researching GPT-3 and am interested about its potential in two apps I’m hoping to develop:

  1. I’m dyslexic; I understand the trouble and complication of being a professional in a world run by text. I have been planning a virtual writing assistant app for dyslexic adult professionals using voice to both generate and edit content (speech to text) and have the app read it back to you (text to speech) powered by the GPT-3 linguistics model… with AI assisting in the process and helping the user to write emails, content, texts, reports, and possibly more long form such as novels.

This would be for people who know what they want to say, just not how to say it. This GPT-3 powered app would help users with that, as simply as possible, with enhancements like thesaurus and dictionary as aids… and the AI helping to create and edit the content.

I see this as a learning tool for anyone (not just the neuro-atypical) but also students, and the English-is-my-second-language professional.

  1. The second planned app is inspired by my career as an independent photojournalist/ editorial photographer: to generate an app that searches the internet for editorial breaking news on a topic, pulling from legitimate sources (e. g. eyewitness social media, local news) and use the linguistic modeling to cross-check for accuracy… and then be able to republish quickly, for public safety announcements, news, etc. without the agenda aspect. Just the facts.

For example: there’s a breaking news story of an earthquake in California, massive amounts of data start appearing on social media; what this app would do is scour the internet for prompt keywords, identify legitimate posts by cross referencing the language of said posts with PSAs, and news stories (such as names of towns, streets, names to verify accuracy of location and data) … like compiling portions of a digital fingerprint. The app would allow you to take all these verified facts, generate a title and repost the collective, verified data in breaking news article form with links (websites, social media, wherever) with the ability to update it as more verified information comes in.

I believe that a large portion of educated society wants and needs direct information, regarding editorial news, and not portions left out which skew or taint the data. An app like this might help shrink the gap between facts and agenda-driven news, and possibly open up a new field in news media. From what I’ve seen, GPT-3 can determine the difference between facts and fictions and can pull from multiple sources with both power and accuracy.

These are my top two app ideas. My questions to this community are: Are either of these ideas viable? Secondly, is there any way to ballpark what it would cost to develop either, or both, of these apps? Thirdly, I am not opposed to having a partner in this venture and am actively reaching out to universities with AI programs to hopefully garner some development partners.

I do have some funds to develop an app but have no idea how to safely hire a legitimate developer / coder that knows anything about working with GPT-3, and I have so many questions on how GPT-3 (and perhaps GPT-4) can be utilized in apps like I’ve described.

3 Likes
2

Hi @steven101 , thanks for sharing this! I am a writing teacher and relatively new to all this also–can’t help with cost estimates or any such practical questions, but I just had a couple of thoughts to share.

Have you looked at the existing writing assistant apps that use GPT-3 like JasperAI? I don’t think they have the speech to text aspect, but they do try to do the rest. The founder of ChibiAI has put a lot on YouTube about his approach.

Your second idea sounds incredibly valuable and reminds me a little of Elicit.org, which is used for scholarly research but must have a way of verifying that might be parallel to your needs.

David Shapiro was really helpful advising people on questions like this a while back–not sure if he is active currently.

1 Like
3

Anna Mills:

Thank you for the reply and your feedback. The first app, which I tongue in cheek have called Cixelsyd, I’m basing on my own experience in college and the work realm, but according to the Yale website, as many as 20% of people show signs of reading challenges, in various forms.
source: Dyslexia FAQ - Yale Dyslexia

What most people like me suffer from is sporadically having words (digital or analog) either turn around in our heads or disguise themselves, or even become invisible. We’re in good company: Edison, Einstein, Spielberg to name a few; most dyslexics are very creative and passionate but are sometimes unable to express it because of some text-based limitations.

In my experience because this linguistic ‘phantom’ exists, most dyslexics lack the confidence to read and write without assistance, which is why the speech-to-text and text-to-speech elements are so pivotal in this fledgling app.

I have seen JasperAI and it just doesn’t meet that need, not that it couldn’t be enhanced to provide that. I just haven’t seen anything yet on the market that checks all the boxes.

Regarding your feedback on the second app, let’s describe it as news with a Digital Fingerprint, I agree that it is not only a useful tool, but might be a requirement moving further into the 21st century’s age of extreme global propaganda.

In a news industry that is already lagging and having to make cuts, it would reduce the ‘journalism behind desks’ aspect and encourage them to be more out in the field. More stories out in the field mean more news for those who just want direct facts. It could revolutionize news as we know it today, and perhaps make a divide between real news and entertainment/ spin.

This would be a successful subscription-based model app for news agencies, or researchers, students et cetera; the GPT-3 based intuitive search engine would be key, with keyword input prompts and using the data to generate fact-based news stories with linguistic fingerprint. Later it could be trained to cross refence editorial images, posted throughout public social media, and connect them to new with a similar digital fingerprint verification.

For instance, if a news story breaks in front of a grocery store parking lot, the app could analyze the photos popping up on social media, triangulate the location, and verify that the photos are indeed connected to the story, for example, no parking lot is ever that same every day, there are so many variables: size of vehicles, colors, positions, tires, wheels, windows, et cetera. The app could likely do this with field photos as well by comparing angles of buildings, tree positions, time of day, positions of the sun, among other things.

I’m a little leery about expounding on these ideas here because I don’t know the community well, and there is the risk of people taking ideas… but I believe in the cause, that these are important apps to get built someday.

I did reach out to David Shapiro and Adam Goldberg, who directed me to this forum. I like it here; it feels properly ‘old school’ for those approaching 50 range, but with new tech. :slight_smile:

Thank you for your time.

1 Like
4

Thank you for explaining–these are both wonderful ideas, and I hope you can build them.

1 Like
5

Hello Steven, welcome.
Shashwat here.

So as I read about your app description of the first app, I’ve already built something similar in the space — a general-purpose chatbot with a speech synthesis feature. And I guess your first idea seems more doable right now than the second one.

I have a couple of beta users, who are using the Chatbot daily, and some have asked for “engineering” their own chatbots according to their use cases. So your app idea roughly falls there IMO, I might launch an enterprise plan to cater to that service.

About the second idea - Now, GPT3 is a text-generation model, it can’t do some of the things you described here and will probably have to use other ML models to make it possible. Like retrieval of images and describing them in plain text, etc.
However, I do think it’s possible to amalgamate technologies and integrate them so they work in sync.

I would invite you to try the chatbot product I’ve built, I assume it’s very close to your first idea. And maybe discuss it further.

Thanks!

1 Like
6

Xerxes: Thank you for the reply and remarks. I would be very interested to try out your speech synthesis bot; just to be clear I’m not a coder. I’m like a car enthusiast that is not a mechanic discussing the fastest engine. But if you think my feedback would be useful, I’d be more than willing to try it.

Regarding the digital fingerprint news app idea, I think it would be a game changer in the space, but yes it would take several partners to build. I do think it would be worth it, and that it’s about time the news industry meets the 21st century.

7

Hey Steven, :grinning:
I’d love to help on this app, i’m working on the same thing and have some good libraries I think would be of use, if you get a moment, let’s chat!

here’s my email: [my email]

1 Like
8

Update to original post 9-21-22

I am looking either to hire a dev/ coder or a partner to develop two GPT-3 enabled apps:

The first is a combo of GPT-3 and DALL-E, a news aggregating/ search app that verifies news stories by cross-referencing useable data, a type of digital fingerprint, especially for time-sensitive news verification and posting. The point is use the vast capabilities of GPT-3 to cross reference data points in the stories and images, to cut down on ‘fake’ or incorrect/ outdated news stories.

The second app is a virtual writing assistant app for professionals with dyslexia–utilizing speech to text and text to speech–as a stand-alone app, enhanced with GPT-3 linguistics model to correct, and enhance user-generated text and read it back to the user, with copy and paste capabilities.

Must be Open AI community registered, approved for codex access and have app dev experience.

1 Like
9

oh, so you’re back after ghosting me on telegram and deleting the chats?
Not a good sign IMO. Beware, guys.

10

Xerxes: I apologize, I have been very busy with business and personal matters.

I tried out your app for a week; I stated to you, directly, that it was very intriguing chatbot, but it wasn’t what I was looking for. Your app didn’t match my scope or my vision for an app for people with dyslexia.

Yes, I deleted your app after I was finished testing it… and gave you lengthy feedback.

I also gave you my phone number and email as the best ways to contact me. If you have any unresolved inquiries, feel free to call or email.