HNHacker News
TopNewBestAskShowJobs

freesam

11 karma · joined June 20, 2022

submissionscomments
freesam··on [dead]
I built a Linkedin automation flow that can

1. Automatically fetch recent news articles based on randomly selected topics

2. Generate thoughtful commentary about the news Connect the news to our company's services

3. Create engaging LinkedIn posts that encourage audience engagement

freesam··on [dead]
We want to share our implementation of the realtime voice-to-voice API with RAG to integrate the AI chatbot with your own knowledge base. You can follow our example to build your own realtime voice-to-voice AI chatbot with your own knowledge base .

https://www.chat-data.com/blog/create-your-own-realtime-voic...

Here is the Youtube video to show how it works:

https://www.youtube.com/watch?v=Ol7X7O9BrCY

You can also expierience it directly here: https://www.chat-data.com/chatbot-iframe/654aee418bc9a33b0e5... 1

freesam··on [dead]
Create a Personalized Experience by Syncing Our AI Chatbot with Your Website!

We’re excited to announce a new feature that enables bidirectional communication between your website and our AI chatbot, enhancing the experience for every visitor!

Here’s how it works:

Seamless User Data Syncing

If your website supports user authentication, you can use the postMessage API to transmit user information to the chatbot upon successful login. This means no more redundant data collection! For example, if you have diagnostic tools generating user-specific metrics, the chatbot can follow up on those insights.

Try It Out!

Visit our Demo Website ( https://openai-assistant-demo.chat-data.com/ ) and see the magic unfold! For instance, if the chatbot knows the user's lucky number is 934383, it can respond accurately to questions like "What’s my name?" or "What’s double my lucky number?" This showcases the power of syncing user data for a tailored experience.

Real-Time Communication

When you embed our chatbot as a widget or iframe, you can receive real-time events such as chat, live-chat-escalation, minimize-widget, and lead-submission. This allows for dynamic interactions that enhance user engagement on your site.

Interested in Implementing This?

Check out our comprehensive documentation ( https://cookbook.chat-data.com/docs/postmessage-events ) on how to get started. We’d love to hear your thoughts or assist you further!

freesam··on [dead]
Here, we use the OpenAI assistant API Math tutor example (https://platform.openai.com/docs/assistants/quickstart ) to create a math tutor.

In the website link ( https://openai-assistant-demo.chat-data.com/ ), you can input your personal information, for example, "My lucky number is 889837"

Then you asked the math tutor, "What is double of my lucky number"

The chatbot will reply

# Defining the lucky number lucky_number = 889837

# Calculating the double of the lucky number double_lucky_number = lucky_number * 2 double_lucky_number The double of your lucky number, 889837, is 1,779,674.

It means that you can directly pass the personal information of your customer to your open AI assitant API to create a personalized experience.

freesam··on [dead]
With the advances of the LLM technology, the ideal customer support flow should be like this

Here is how the work flow works:

1. The customer interacts with the AI chatbot to receive responses on the website.

2. The AI chatbot can instantly answer most(suppose 80%) of queries. However, some corner cases ( suppose 20%) may require human intervention.

3. In such cases, the customer can click the "live chat escalation" button to request human support. The human agent receives an email notification with a unique handling URL. This

4. URL is whitelabeled, allowing the agent to handle the query without logging into Chat Data.

5. If the customer is still on the chat widget, the agent and the customer can communicate in real-time, similar to a chat room.

6. If the customer has left the chat widget interface, an email with a unique handling URL will be sent to remind the customer of the new message from the human agent. This email can be sent from a whitelabeled email address.

7. When the customer responds to the chat message, the human agent receives an email notification about the new messages. All emails and handling URLs can be whitelabeled and do not require a login to manage the messages.

Either human or AI chatbot alone is not the best solution for the best customer support experieince

freesam··on GPT-4o is consistently hallucinating
We found that gpt-4o is consistently hallucinating. Here is an example: [ https://platform.openai.com/playground/chat?models=gpt-4o&pr... ] In this case, we never mentioned any store location in Ringsted. However , gpt-4o still fakes a store in the address: Adresse: Ringstedet, Klosterparks Allé 10, 4100 Ringsted Åbningstider: Mandag-fredag: 9.30-19, Lørdag: 10-17, Søndag: 11-16 Tlf: 50603398 Email: ringsted4100@gmail.com This hallucination is quite consistent. The gpt-4-turbo or even gpt-3.5 model doesn't have the same issue and correctly acknolwedged that there is no store in this location.
freesam··on [dead]
OpenAI's Assistant API is becoming more and more power now. Maximize your Assistant API's potential by deploying it across multiple platforms—your website, Discord, Slack, and WhatsApp—in just 20 minutes with minimal code. We've open-sourced the code how to integrate your Assistant API with all the the four platforms with only a few lines of code within 20 mins.
freesam··on [dead]
AI content is everywhere. Don't wanna people to feel nauseated with your AI generated website and Google to crack down on your ranking. Try this tool to make your website more like human-written and remove the AI footprints of your content.
freesam··on [dead]
I sponsored the 1st OpenAI API credits usage competition with $1000 in this twitter ( https://x.com/ysu_ChatData/status/1774126917714903126?s=20 ) Lots of people love it. However, $1,000 OpenAI credits get used up in just one day and the people who needs the OpenAI credits cannot use the API key I gave for long enough time. Lots of people still love to use the free API key I provide to build something. Therefore, I plan to sponsor another $1000 for another 2-week OpenAI API credits usage competition with better rules.

Competition Rules:

1⃣ Share your project plans, model details, and estimated token usage like the example in the discord channel.

2⃣ Keep your API key to yourself. No sharing!

3⃣ Report back on your project outcomes.

Max $30 consumption per project.

Competition ends on April 30th or when the allocation of $1000 OpenAI credits are consumed.

Share your progress and support each other!

@ysu_ChatData will update daily on API key consumption.

Repost this and join our Discord channel.

Everyone can view progress, but API keys are only given to those who repost, share their projects. Let @ysu_ChatData know by DM if I missed you!

"Ready to innovate? Join https://discord.gg/MGTWxeX2?event=1224476762952568904 now to secure your spot! Once the estimated credit usage surpasses $1500 ($500 as buffer), we'll NOT include more participants. This ensures everyone has ample OpenAI credits for the competition. Limited spots available! Here is the server https://discord.gg/JaAa8mqW

freesam··on [dead]
Rules:

1. Reply to and repost this tweet.

2. You agree that you can only use the key for yourself.

3. The competition ends in 2 weeks.

$1000 of openai credits are allocated

freesam··on [dead]
Hi all, We used this self-made API to get the traffic of 200 Shopify stores within 1 minutes. In production, we are able to get 1000 high quality Shopify stores from a list of 10,000 shopify stores. Without this script , the mission will be impossible. The script is open sourced at https://github.com/chat-data-llc/shopify_store_traffic_api . If you are working on a similar task, it should be helpful to you. Totally free.
freesam··on [dead]
Since the GPTs store is so popular , why not open the Shopify store in the GPTs store so that people can buy and sell the products in the GPTs . I made a demo to proof the concept.
freesam··on We crawled 800M PubMed articles and made this GPTs
Problems:

1. There is no GPTs that really have access to the PubMed articles database but mostly use the ChatGPT + Google Search to do the work. Or use the Search method of the PubMed to do the work. 2. Those GPTs uses only the search results but never looked at the content of the PubMed article to provide answer

Therefore, the results of their GPTs will be really bad.

Solution:

This GPTs used all the content of the 800 Million PubMed article to create a distributed database and use the RAG pipeline to serve the articles. Because the RAG pipeline understand the content of the article, it can provide the most relavant PubMed article for serious research purpose

freesam··on Medical Question-Answer AI Model Evaluation Framework
I've worked with a few medical LLMs, and their depth of knowledge and holistic interactions (knowledge breadth) are absolutely incredible. My gut instinct is that this will help fill in the massive shortage of GP physicians (other examples: NPs, PAs), and perhaps heavily assist or replace pharmacologists.
freesam··on [dead]
Discover the power of personalized chatbots for your business, no coding required! Unlock privacy assurance, tailored solutions, customization, insights, and flexible deployment. Learn step-by-step methods using platforms like Chat Data for confident navigation of market challenges.
freesam··on Medical Chat Model Sets Unprecedented Benchmark in Medical Question-Answering
Medical Chat was evaluated on MedQA, a benchmark encompassing a diverse dataset from various medical board examinations. The model achieved an outstanding accuracy rate of 97.8%, securing the top position on the Official Leaderboard.
freesam··on Medical Chat stands out as the most accurate medical question-answering system
Medical Chat model has achieved an impressive 98.1% accuracy in USMLE and 97.8% in MedQA Standard Question-Answering Tests, surpassing Google's Med-PaLM as the most accurate medical question-answering system for public use.
freesam··on [dead]
This AI is solely trained based on the blog of Sam Altman: https://blog.samaltman.com/ . Ask me anything about what Sam has said
freesam··on [dead]
Chat Data is a flexible platform that enables you to create AI chatbots using your chosen data, custom backend, or pre-trained models provided by us. It offers flexibility in both the frontend and backend, along with a comprehensive user analysis tool.
freesam··on Craft your own AI customer support bot with one-line of code
use case: Build a AI customer support to any websites and answer questions specifically about your websites
freesam··on Peer-to-peer mock interview matching platform
PairCoding is a platform that match peers with similar mock interview practice needs to practice interviews together. For example, peers can target the same company for the same type of interviews. After the interview, each person can rate the other peer from the interviewer's perspective, which helps improve the interview skills.

Our platform will provide questions (real interview questions categorized by company), hint and example solution, video chat experience, code compilation & execution, feedback submission in the interview experience. Even a new graduate could act as an experienced interviewer in our platform.