You've seen the hype. Now you're curious. Why not have a crack at AI using this online lab...

...from Microsoft, which also just bought upstart Bons.ai


Microsoft has thrown open the doors to its AI Lab, a suite of beginner projects to help developers learn machine learning.

There are five different experiments that cover computer vision, natural language processing, and drones. “Each lab gives you access to the experimentation playground, source code on GitHub, a crisp developer-friendly video, and insights into the underlying business problem and solution,” according to Microsoftie Tara Shankar Jana on Tuesday.

The first one is is the DrawingBot. It teaches developers about generative adversarial networks (GANs), a popular type of neural network that learns to create similar content to the data it was trained on. DrawingBot uses the AttnGAN - short for attentional GAN - created by a group of researchers from Microsoft, Lehigh University, Rutgers University and Duke University.

doctor_scan

An AI a day keeps the doctor away... Neural net software gets better at clocking cancer tumors

READ MORE

AttnGAN generates an output image based on input text. DrawingBot focuses specifically on birds, and inspects the individual words in the inputs and learns to map the description to a specific region of picture of a bird.

For example, it will learn to pick up things like ‘red wings’ or ‘orange beak’ to build up the image. If the description sways too far from what it was trained on then AttnGAN can’t quite match it. So, it’s pretty good at generating realistic looking images of birds, but not ones with, say, two heads or gigantic dragon-like wingspans.

JFK Files is a less flashy. It uses Microsoft’s Cognitive Search and Azure Search together and applies natural language processing to search through documents for relevant information.

In the AI Labs example, Microsoft uses the files on the JFK investigation that have been declassified and are now public to demonstrate how it all works. It can pull things like names and dates out of the old documents, some of which are handwritten.

The third one on the list is style transfer. It uses the popular COCO dataset with over 200,000 labelled images. The objects in the photos are segmented and carried over when the style transfer is applied. It turns realistic photographs into mosaic-style images, or in a surrealist style ones.

It uses Microsoft’s Visual Studio Tools to train the model and developers can play around with the code written in Keras and TensorFlow and execute it on Azure Cloud using Nvidia GPUs.

The fourth is about text understanding. Machine Reading Comprehension uses the popular SQUAD dataset to provide short paragraphs for a model to read and it can answer questions. Microsoft uses its Reasoning Network (ReasoNet) to do this and hopes it can be used for enterprise data and to help customer service by answering any questions a customer might have about a specific application - a bit like FAQs.

Finally, there is the simulated drone environment AirSim. The goal is to pilot a drone around a soccer field that is littered with stuffed animals. Users have to write a program in Python to fly the drone around and identify all the animals.

The code can be run on a real drone. It needs to be exported to TensorFlow and then sent to Docker containers, before its deployed using Microsoft’s Azure IoT Edge platform run on a drone using an Nvidia GPU.

You can play with them all here. ®


Other stories you might like

  • Microsoft promises to tighten access to AI it now deems too risky for some devs
    Deep-fake voices, face recognition, emotion, age and gender prediction ... A toolbox of theoretical tech tyranny

    Microsoft has pledged to clamp down on access to AI tools designed to predict emotions, gender, and age from images, and will restrict the usage of its facial recognition and generative audio models in Azure.

    The Windows giant made the promise on Tuesday while also sharing its so-called Responsible AI Standard, a document [PDF] in which the US corporation vowed to minimize any harm inflicted by its machine-learning software. This pledge included assurances that the biz will assess the impact of its technologies, document models' data and capabilities, and enforce stricter use guidelines.

    This is needed because – and let's just check the notes here – there are apparently not enough laws yet regulating machine-learning technology use. Thus, in the absence of this legislation, Microsoft will just have to force itself to do the right thing.

    Continue reading
  • Is computer vision the cure for school shootings? Likely not
    Gun-detecting AI outfits want to help while root causes need tackling

    Comment More than 250 mass shootings have occurred in the US so far this year, and AI advocates think they have the solution. Not gun control, but better tech, unsurprisingly.

    Machine-learning biz Kogniz announced on Tuesday it was adding a ready-to-deploy gun detection model to its computer-vision platform. The system, we're told, can detect guns seen by security cameras and send notifications to those at risk, notifying police, locking down buildings, and performing other security tasks. 

    In addition to spotting firearms, Kogniz uses its other computer-vision modules to notice unusual behavior, such as children sprinting down hallways or someone climbing in through a window, which could indicate an active shooter.

    Continue reading
  • Cerebras sets record for 'largest AI model' on a single chip
    Plus: Yandex releases 100-billion-parameter language model for free, and more

    In brief US hardware startup Cerebras claims to have trained the largest AI model on a single device powered by the world's largest Wafer Scale Engine 2 chip the size of a plate.

    "Using the Cerebras Software Platform (CSoft), our customers can easily train state-of-the-art GPT language models (such as GPT-3 and GPT-J) with up to 20 billion parameters on a single CS-2 system," the company claimed this week. "Running on a single CS-2, these models take minutes to set up and users can quickly move between models with just a few keystrokes."

    The CS-2 packs a whopping 850,000 cores, and has 40GB of on-chip memory capable of reaching 20 PB/sec memory bandwidth. The specs on other types of AI accelerators and GPUs pale in comparison, meaning machine learning engineers have to train huge AI models with billions of parameters across more servers.

    Continue reading

Biting the hand that feeds IT © 1998–2022