Ivan Zhou – Analyst, Global Industry
A Glimpse into Google’s AI Efforts at Google I/O
Google I/O has become an annual festival, where the tech giant shows off its AI muscles and imagination with a series of new product announcements. AI-related presentations at Google I/O 2019 ranged from infrastructure to products that regular consumers can interact with on a daily basis. On the infrastructure side, Google announced powerful new Cloud TPU Pods that can train a ResNet-50 model in 7 minutes and BERT in 76 minutes. This is ideal for fast model iteration and frequent retraining. Google also announced updates to its flagship AI product Google Assistant, which is running on one billion devices globally. The new generation Assistant will be able to run locally with a smaller model size and carry out a number of tasks simultaneously, and adds more personalization features to serve users better. There were also updates to Google Duplex, Pixel phones, Lens, and other AI products. Since the Google I/O last year, we have seen a clear trend with Google integrating AI deeply into infrastructure and products to create value. This is definitely healthy and constructive for the AI ecosystem.
May 7 – Google Announces New Cloud TPU Pods
- Google announced the v2 and v3 models of its cloud-based supercomputers Cloud TPU Pods in beta.
- The v2 Pods feature up to 512 core and 256 TPU chips, while v3 Pods feature up to 2048 cores and 1024 TPUs. Each TPU in v3 pods can compute at the speed of 100 petaFLOPs. Both machines can be rented at unit level or slices of one pod.
- With a v3 Pod, you can train a ResNet-50 model in 7.1 minutes and BERT in 76 minutes. TPU pods are designed for rapid model prototyping, fast training iteration, and frequent model update.
May 7 – Google Assistant will Add Multi-task Support and “Picks for You”
- Google Assistant will be able to be stored locally on a phone by reducing the size of the voice recognition models from hundreds of gigabytes to around 500 MB. It will be able to run faster without latency to connect to the cloud and work offline.
- Google Assistant will be able to take a series of requests after the waken command “Hey Google”. For example, a user can ask the Assistant to write a text message, search for a phone, add the photo to the message, and send it without repeating the “Hey Google” command.
- Google Assistant will also launch a personalization feature “Picks for You” that replies with recipes, podcasts, and events tailored to users’ preferences.
Robert Tian — Senior Analyst, Global Industry
Siemens Enters 2nd Round of Vision 2020+
Last year Siemens released plans for a simplified and leaner company structure with Vision 2020+. In its 2019 annual midterm review, Siemens announced it has completed the first step of Vision 2020+ and that its medium-term digital business targets were fulfilled. The next phase of Vision 2020+ has been approved by the Siemens AG group board of directors and will leverage AI, IoT, and cutting-edge technology to hone the company’s business model and operations in digital industries and smart infrastructure fields.
May 9 – HP and Siemens Deepen Additive Manufacturing Alliance to Advance Digital Manufacturing
- Siemens partnered up with HP in AI-based 3D printing system development in industry-level manufacturing.
- The companies look forward to developing an automated 3D printing solution to enable unique product designs and digital factory possibilities.
May 7 – Siemens to Build Focused Energy Powerhouse and Further Boost Performance
- The board of Siemens AG approved the next steps in the Vision 2020+ strategy after the company completed its medium-term growth and reached profit targets.
- The company will focus on digital industries and smart infrastructure and predicts the operations will save US$2.2 billion by 2023.
10 AI News You Must Know
May 9 – Cisco open sources MindMeld conversational AI platform
May 9 – India Mutual Funds Asked to Submit Quarterly Report on AI, ML Use
May 8 – Air Canada Forms Artificial Intelligence Labs
May 7 – Square Acquires Conversational AI startup Eloquent Labs
May 6 – G-7 Science Academies Release Statements on Science and Trust, Artificial Intelligence, Citizen Science
May 3 – Ericsson Opens Artificial Intelligence Lab in Montreal
May 2 – Samsung Expands Is Artificial Intelligence Lab in Canada
May 1 – IBM, Thomson Reuters Introduce Powerful New AI and Data Combination to Simplify How Financial Institutions Tackle Regulatory Compliance Challenges
April 30 – AstraZeneca Signs on BenevolentAI for Lung, Kidney Drug Discovery Work
April 30 – UiPath Nabs $568M at a $7B Valuation to Bring Robotic Process Automation to the Front Office
Presented by Synced Industry Analyst Team