Top AI Trends at Google IO Connect Conference

In the next couple years, AI will impact our workforce in profound ways. Are we ready? At the Google IO Connect conference in Miami, FL, community leaders and team members from Google gathered to celebrate some of the new advances in AI tools, web, mobile, and cloud computing. The experience at Google IO Connect and the related leadership conference still has me excited. I want to thank Google for investing in GDG leaders, student groups, and Women Tech Makers. Loved the theme of working together in our local area to make a difference.

New players like OpenAI/ChatGPT have engaged our industry in exciting discussions of the value of large language models and conversational AI. As an industry, we have discussions regarding how we train machine learning models while respecting the data rights of data set creators. We’re asking hard questions around making responsible AI and bot systems that promote safety.

At the Google IO Connect conference, Google emphasized their efforts to promote responsible AI. We openly discussed our collective responsibility of fostering equity, fairness, and respecting intellectual property in our developer ecosystems.

While AI/ML will create many jobs, this technology will also displace jobs as well. In my view, we have a responsibility as technology makers to help our communities pivot in the face of this economic displacement. For this reason, I really enjoyed meeting like minded community leaders cross Google developer group (GDG) network, Google developer student groups (GDSC), Women tech makers (WTM), and Google developer experts (GDE). Community makers also have a key role in helping communities connect to new economic opportunity. As influencers, we care about empowering our communities with learning, inspiration, and community to adapt to a deep economic shift that will continue. Some say that robots and AI will come for our jobs. We believe that people matter first! We believe that we can inspire more people to become new forms of creators, makers and developers. How do we help our communities connect to new classes of jobs that do not even exist yet?

For our community, I wanted to share a few favorite AI items from the conference.

Popsign: I appreciated work between Google and Georgia tech to empower people to learn sign language. Their tool created a gamified experience to learn sign language in a novel manner. This tool feels like DUO lingo for sign language. For many people, this tool changes everything. In theory, if developers want to extend this technology, the machine learning model has become open source to encourage further innovation and extension.

Bard, Palm API and MakerStudio: It’s becoming hard to avoid the ChatGPT vibe and large language models. (LLM) We have enjoyed seeing Google unpack Google Bard, an AI conversational bot with comparable capabilities as ChatGPT. In the early Internet, the browser wars fueled innovation to improve web experience that we continue to enjoy today. It’s great to see that ChatGPT may have some large-scale competition from Google and open-source alternatives. Prompt engineering requires an artful kind of thinking. Now we can use English for a completely novel form of computation and reasoning. I appreciate that MakerStudio tries to guide developers through the concepts of prompt development from simple to complex. I can say that the Palm API feels very accessible so far. Looking forward to exploring Palm API and the related MakerStudio tools more extensively. Make sure to play with MakerStudio below.

https://makersuite.google.com/

Tools to find data sets and machine learning models: ¬If you seek a great index of open-source code, we turn to Github. Where do you find pre-trained high value machine learning models? Where do you find machine learning models by task or problem domain? How do you find quality data to enable your machine learning research? As an industry, we now have several interesting repositories:

AI image generation: From the TensorFlow/Keras team, Google has worked to put complicated AI operations into a concise style with lots of power. For a seasoned Linux user, it might take you a good week or two to setup the tools to execute an AI image generation job. (Aka. Stable diffusion) This Python “colab” notebook provides a nice overview of doing AI image generation in a small form factor. I created a silly ninja playing a guitar image in less than 10 minutes using KerasCV.

https://colab.research.google.com/github/keras-team/keras-io/blob/master/guides/ipynb/keras_cv/generate_images_with_stable_diffusion.ipynb

In a similar manner, I look forward to digging into the new API tools for natural language processing with KerasNLP. Under the Keras tools, Google has encouraged concise and high impact set of APIs ranging from language processing, GPT2, and computer vision.

WebML: As a web developer, I enjoy seeing how our modern web browsers can execute amazing machine learning tasks disconnected from cloud providers. In the past, you needed a Kinect sensor to detect human body pose. With TensorflowJs, developers can sense body or hand movements using pre-trained models directly in the browser. With advances in WebGPU, I can imagine the power of machine learning in the browser will continue to grow. I can only imagine the advances in object detection and scene understanding we will see in the next five years. I also love the application of TensorFlowJS in music.

Be the first to comment

Leave a Reply

Your email address will not be published.


*