Subscribe
Sign in
NLP inside your database — Query OpenAI’s GPT-3 from directly inside your database
You’ve explored the power of GPT-3 & ChatGPT; now you can apply that power to your own data by bringing GPT-3 to your database with MindsDB, to deliver additional insights & value to your existing data.
MindsDB is an Open-Source ML Platform for Developers
353
Replies
Hey Product Hunters!
Adam here, co-founder of MindsDB, the fastest growing open source applied machine learning platform in the world 🙂
Today we are excited to launch our NLP integration with Open AI’s GPT-3 and Huggingface pre-trained models, bringing the power of two of the world's most popular NLP frameworks to your database. This means you can now harness the power of these models on your own data directly inside your existing database with only a few queries.
We believe that all applications will & should be enhanced or powered by AI, but right now, this is painful and slow. The workhorse that is the full-stack developer sets up the database, builds the front end and calls for the API, but yet if they want to incorporate even the simplest Machine Learning into the process this requires a team of specialists, one that may not even be available in some companies.
Thus at MindsDB, we want to empower the full-stack developer to add a new pillar to their tool kit….. Machine Learning, and with MindsDB this is now possible.
MindsDB is an Open-Source Machine Learning Platform for Full Stack Developers. We currently integrate ~70 popular databases with state-of-the-art machine learning frameworks, and this number is constantly growing thanks to the amazing open-source community!
So, here’s some of the things you can do (with just a few SQL commands 😲) all in your DB:
Classify and label rich text, for instance, for sentiment analysis, detecting hate speech, or spam;
Extract meaning for labelling text even when you don't have any training data - so-called zero-shot classification;
Have conversations, answer questions or comments;
Automatically summarize long texts, translate them, and more.
Upcoming Webinar: We are also hosting a deep dive webinar on our NLP capabilities, including GPT-3 and Huggingface. Sign-up here
Check out this Demo Video and Get Started with MindsDB.
We would love your feedback.
Kind regards,
Adam
I love MindsDB and can't recommend the service more highly. We have used MindsDB Cloud to integrate ML within our product effortlessly. The team is incredibly supportive and helpful too. Congrats on the launch!
Looks really cool. Curious what kinds of validation you have on the output?
One of the things keeping me from using GPT in production is its tendency to "hallucinate".
@stedmanblake One thing that can certainly help is the 'temperature' parameter. This is enabled with MindsDB - you can set it low to make sure the answer is at least more deterministic
https://platform.openai.com/docs...
@stedmanblake Indeed. Solving these hallucinations is an active research area. In theory, architectures like DeepMind's RETRO should be better at this. MindsDB's approach could prove very useful here by providing a simple and fast integration with knowledge bases via our DB handlers.
@stedmanblake this is a very good point, I think it would be very cool to measure, certainly we should be able to ask the model if the info is made up or not, i will make some research on this, would you like to collaborate?
@stedmanblake MindsDB takes the accuracy and reliability of its predictions very seriously and implements a number of validation techniques to ensure that the output is trustworthy. Some of the validation techniques used by MindsDB include:
1. Data quality checks: MindsDB performs data quality checks on the input data to ensure that it is clean and suitable for model training. This helps to prevent issues with the output that can arise from poor-quality data.
2. Model performance evaluation: MindsDB evaluates the performance of its models on a validation dataset to ensure that they are performing well and not overfitting to the training data. This helps to prevent issues with the output that can arise from models that are too complex or not well-suited to the data.
3. Model interpretability: MindsDB provides interpretability features, such as feature importance, that allow you to understand how the model is making its predictions. This helps to ensure that the output is reasonable and that the model is not relying on obscure or unusual relationships in the data.
4. Post-prediction analysis: MindsDB provides tools for post-prediction analysis that allow you to examine the output of your models and make sure that it is accurate and makes sense. This helps to identify any issues with the output and ensure that the model is performing as expected.
Overall, MindsDB takes a comprehensive approach to ensure the accuracy and reliability of its predictions and provides multiple tools and features to help you validate the output and ensure that it meets your needs.
Well done launching such a richly featured product 🔥!
I was wondering - do you allow users to train a model against existing data in one's database? And as part of the training do you allow them modify parameters etc?
@ed_forson absolutely! please join our community on slack, Patricio Cerda from our team, has implemented the ability to finetune based on your data.
@ed_forson Yes, MindsDB allows users to train models against existing data in their database. The training process involves feeding the data into the MindsDB engine, which uses machine learning algorithms to learn the relationships between the input features and the target variables. During the training process, users have the option to modify various parameters, such as the learning rate, the number of trees, the maximum depth of the trees, and more, to optimize the performance of the model.
In addition, MindsDB provides a user-friendly interface that makes it easy for users to train, test, and deploy their models. The interface allows users to visualize their data, set their target variables, and perform feature engineering tasks like normalizing and transforming their data. Once the model is trained, users can evaluate its performance using metrics like accuracy, precision, recall, and F1 score, and make further adjustments as needed.
Overall, MindsDB is designed to be a flexible and user-friendly tool for building and deploying AI models and allows users to train models against their own data and adjust various parameters to optimize performance.
@pradeeb28 You can either deploy it on your own servers (like AWS) or you can start by testing it out on our cloud (https://cloud.mindsdb.com/home). We also have a quickstart on our docs here which is a great place to start. (https://docs.mindsdb.com/quickstart), and if you need any help our community is always around for support. (https://mindsdb.com/joincommunity)
@pradeeb28 Also, you will be able to integrate MindsDB directly with Supabase https://docs.mindsdb.com/data-in....
@pradeeb28 we are also working on SDKs, if you join our community on slack we can talk about the current SDK implementation.
@pradeeb28 If you're proficient in building UI and use Supabase as a backend for basic CRUD, getting started with MindsDB should be straightforward. Here are the steps you could follow to get started:
1. Sign up for MindsDB: You can sign up for a free account on the MindsDB website to start using the platform.
2. Explore the documentation: MindsDB provides comprehensive documentation that explains how to use the platform, including how to train models, make predictions, and access other features.
3. Connect to your data: MindsDB supports multiple data sources, including SQL databases like Supabase. You can connect to your Supabase database and import your data into MindsDB to start training models.
4. Train a model: Once you have your data in MindsDB, you can start training models by specifying the target column and selecting the features you want to use. MindsDB will automatically select the best model for your data and train it for you.
5. Make predictions: After your model is trained, you can start making predictions by sending new data to MindsDB. You can use the MindsDB API or the JavaScript SDK to make predictions from your front-end application.
6. Evaluate the output: Finally, you can evaluate the output of your models to ensure that they are accurate and make reasonable predictions. MindsDB provides tools for post-prediction analysis that allow you to examine the output and make sure that it meets your needs.
Overall, the process of getting started with MindsDB is straightforward, and the platform provides many tools and features to help you train models and make accurate predictions quickly and easily.
Congratulations on the launch of MindsDB - I'm excited to see what insights and value it can bring to my existing data!
@usamaejaz thank you!! join our community on slack, we would be looking forward to hear about your discoveries.
@usamaejaz Thanks a lot. MindsDB community will always be there to help you in your ML journey.
Hey @adam_carrigan ! I loved MindsDB and I saw that you integrated with Airtable which I use a lot in my company. Could you provide some more details about what tasks can be handled? I'm interested, but I don't want to book a discovery/demo call if I'm not sure it will be applicable to me. Thanks in advance!
@Adam Carrigan and @Rashid Azarang Esfandiari, I am a fan of Airtable as well. Our docs provide a comprehensive list of tutorials and examples for NLP-based text tasks, including NLP strategies. In addition, we offer other strategies such as regression, time series, and classification. I encourage you to proceed with scheduling the discovery call. We look forward to hearing about your use case.
@adam_carrigan @rashid_azarang_esfandiari if you join our community we can help you there, but please check the integration docs here: https://medium.com/@prathikkshet...
@rashid_azarang_esfandiari MindsDB's integration with Airtable allows you to use the power of AI to automate a wide range of tasks, such as:
1. Predictive modeling: You can use MindsDB to train models that make predictions based on your Airtable data. For example, you could train a model to predict customer churn, sales forecast, or lead conversion rate.
2. Data classification: You can use MindsDB to classify your Airtable data into different categories based on specific criteria. For example, you could use MindsDB to classify customer feedback into positive, neutral, or negative categories.
3. Data imputation: You can use MindsDB to fill in missing values in your Airtable data. This is useful when you have incomplete data that you need to complete for analysis or reporting purposes.
4. Anomaly detection: You can use MindsDB to identify outliers and anomalies in your Airtable data. This is useful for detecting potential fraud, identifying data quality issues, or detecting unusual behavior in your data.
5. Time series forecasting: You can use MindsDB to make predictions about future values in time series data stored in Airtable. For example, you could use MindsDB to forecast sales for the next quarter based on historical data.
6. Recommendation systems: You can use MindsDB to build recommendation systems that suggest items or actions based on your Airtable data. For example, you could use MindsDB to recommend products to customers based on their purchase history.
Overall, MindsDB's integration with Airtable provides a powerful tool for automating a wide range of tasks using AI. Whether you're working with time series data, customer data, or any other type of data, MindsDB can help you make more informed decisions, save time, and improve the accuracy of your predictions.
Great, congratulations on your launch! Good luck to you. I'm already using GPT-3 in SEO.
@new_user_273dccefac @ianu82 I think on keywords research, workflows on SEO, Product descriptions ETC.
This is awesome - I also found the demo very clear and concise as a non-technical person! Well done MindsDB team!
@pulkit_khanna great to hear! would love to see you in our community channels on slack!
Very promising project with cloud and community version 💪. We are already looking into it. Deserves an upvote.
@telecomsxchange Thank you, our team is available to help should you have any questions.
Really awesome! Do you allow closures such that I can ask GPT to write an SQL request for me? :-)
@afchavezt You can create a free demo MindsDB account (https://cloud.mindsdb.com/) to try it out yourself!
@realmaxrush Thanks, Max, would love to hear about how you use MindsDB. Always LOVE hearing a new use case.
good job team for this awesome product.
congrats @jorge_torres2 @adam_carrigan
@kovid_batra Feel free to join our community for discussions and support! https://mindsdb.com/joincommunity
Snoopstein