Jonathan O'Neill

Software Developer / Architect

314 words

Google held their annual developer conference Google I/O last week form May 17th to 19th. There were no hardware announcements at the event so no Pixel phones, Chromebooks or other gadgets. The most interesting thing about the keynote was not the new products and feature announced but the emphasis on AI (Artificial Intelligence) and that Google sees itself as transitioning to being an “AI first” company.

The product features they did announce built on the AI theme like wider availability of Google Assistant on more Android devices and iOS. Assistant was launched with the Google Home smart speaker in the US in 2016. Google Lens will also be added to the assistant later in the year bringing an image input in addition to voice.

Google Photos will get additional features like suggested sharing and shared libraries based on AI which will suggest who to share photos with based on who is in the picture (that is the magic bit).

In order to build these features into its products Google has needed to build the enabling hardware and software stack. The good news is that much of this is been added to the Google CloudPlatform for third parties to leverage the Google technology for their own products or research.

There is a new site google.ai as a home for information on what is available:

  • Cloud TPUs - Googles Tensor Processing Units (TPU), these are custom processing units built to accelerate both training and running of Machine Learning (ML) models.
  • Cloud Machine Learning Services - provides pre-trained models for video analysis, image analysis, speech recognition, natural language analysis and translation.

The wider availability and falling cost of these specialist Machine Learning APIs and compute infrastructure from Google and competing offerings from Amazon and Microsoft are pushing the state of the art in the fields of AI and ML and opening them up to a wider community.

You may also enjoy

Trustworthy AI

1125 words

We are already using AI models whether we are conscious of it or not. Models are increasingly being embedded in all sorts of applications including those making high impact decisions. Trusting these models will be a key to their adoption.

Image Processing with Torchvision

1252 words

This is a brief overview of the most common image-processing tasks that can be performed using the PyTorch Torchvision package. Including classification, localisation, object detection, instance segmentation and semantic segmentation.

COVID-19 - Ireland Cases

478 words

Download Covid-19 case data from Ireland’s COVID-19 Data Hub and tabulate and plot using a Python script.

Data Logger - Part 3 - Visualization

586 words

This is the third in a series of posts to build a simple data logger system for temperature and relative humidity. Part 3 will use the Python Pandas package to visualize the data logged in the Amazon Timestream database.