Welcome to Nural's newsletter focusing on how AI is being used to tackle global grand challenges.
Packed inside we have
- FarmVibes: Microsoft open sources its ‘farm of the future’ toolkit
- AI generated conversation between Steve Jobs & Joe Rogan
- and AI-generated imagery is the new clip art as Microsoft adds DALL-E to its Office suite
If you would like to support our continued work from £1 then click here!
Marcel Hedman
Key Recent Developments
FarmVibes: Microsoft open sources its ‘farm of the future’ toolkit
Jake Siegel
What: Microsoft recently open sourced FarmVibes.AI, a collection of artificial intelligence models that farm operators can use to perform tasks such as planting crops more efficiently. The key hypothesis is that merging a variety of data sources can help create the ultimate truth about a farm. Typical data fused together include, satellite imagery (RGB, SAR, multispectral, different resolutions), drone imagery (RGB, multispec, hyperspec), weather data (historic, current, and forecasts), sensor timeseries data (soil as well as atmospheric), and more (e.g., information from industrial vehicles such as tractors).
Key Takeaway: Bringing ML to the farming industry offers great climate and efficiency potential. Doing this via an opensource toolkit provides the perfect playground for builders to create tailored tools to support farmers (e.g. with harvest date prediction, climate impact estimation, crop identification or micro climate prediction).
CircularNet: Reducing waste with Machine Learning
What: While recycling is a critical action to support with maintaining climate health, the majority of people find the act extremely confusing to navigate. As a result, we do a poor job of recycling right, with less than 10% of our global resources recycled, and tossing 1 of every 5 items (~17%) in a recycling bin that shouldn’t be there. For those that make it to a material recovery facility, the exceptionally cluttered and contaminated nature of the waste stream makes automated waste detection challenging to achieve, and the recycling rates and the profit margins stay at undesirably low levels.
The CircularNet model has been created to support the detection of in material recovery facilities. It performs Instance Segmentation by training on thousands of images with the Mask R-CNN algorithm.
Key Takeaway: The model offers a great foundation for continued work to support recycling efforts at the material recovery facilities. Already the model has the ability to detect individual instances as well as key characteristics about that instance (see image below)
AI-generated imagery is the new clip art as Microsoft adds DALL-E to its Office suite
What: "Microsoft is adding AI-generated art to its suite of Office software with a new app named Microsoft Designer.
The app functions the same way as AI text-to-image models like DALL-E and Stable Diffusion, letting users type prompts to “instantly generate a variety of designs with minimal effort.” Microsoft says Designer can be used to create everything from greeting cards and social media posts to illustrations for PowerPoint presentations and logos for businesses.
Microsoft is also adding an AI text-to-image model to its search engine Bing."
Key Takeaway: The explosion of image generation models has been a recurring theme over the past few months. Therefore, it is good to see these use cases being embedded in real applications instead of remaining as PoCs. As Microsoft Designer has been built upon OpenAI's DALL-E 2, there is an outstanding question on whether any rights should be attributed back to the creators of the publicly soured art that formed the training dataset.
AI Ethics
🚀 AI scans Hurricane Ian damage to give cash to poorest victims
🚀 This podcast brings Steve Jobs back to life, thanks to AI
🚀 Mortgage Fraud AI Improves Detection By 30%
🚀 AI Language Models Are Struggling to “Get” Math
Other interesting reads
🚀 UL2 20B: An Open Source Unified Language Learner - a novel language pre-training paradigm that improves the performance of language models universally across datasets and setups [Google AI]
🚀A Harvard Framework for Generating Research Ideas
🚀 Mintaka: A Complex, Natural, and Multilingual Dataset for End-to-End Question Answering [Amazon]
🚀 Mind's Eye: Grounded Language Model Reasoning through Simulation - Current language models are still challenged by simple questions that require a good understanding of the physical world.
Cool companies found this week
Text to video
Phenaki - A model for generating videos from text, with prompts that can change over time, and videos that can be as long as multiple minutes.
Climate
Farmerline - A Ghana-based company utilizing AI to provide information and resources for farmers, raised $1.5m in Seed funding
and finally...
AI generated conversation between Steve Jobs & Joe Rogan
AI/ML must knows
Foundation Models - any model trained on broad data at scale that can be fine-tuned to a wide range of downstream tasks. Examples include BERT and GPT-3. (See also Transfer Learning)
Few shot learning - Supervised learning using only a small dataset to master the task.
Transfer Learning - Reusing parts or all of a model designed for one task on a new task with the aim of reducing training time and improving performance.
Generative adversarial network - Generative models that create new data instances that resemble your training data. They can be used to generate fake images.
Deep Learning - Deep learning is a form of machine learning based on artificial neural networks.
Best,
Marcel Hedman
Nural Research Founder
www.nural.cc
If this has been interesting, share it with a friend who will find it equally valuable. If you are not already a subscriber, then subscribe here.
If you are enjoying this content and would like to support the work financially then you can amend your plan here from £1/month!