Google Cloud Next '23

Google Cloud will welcome a lot of people to San Francisco for the first in-person event. The CEO of Google Cloud is feeling excited to bring several customers & partners together to showcase the outstanding innovations. It has been a long time since they have been working across a complete infrastructure portfolio, Data & AI, Workspace Collaboration, and cybersecurity solutions.

 Google Cloud

Google Cloud has achieved a few milestones, like reaching a $32B annual revenue run rate in Q2 2023 event. They have also shared a few awesome stories about how they work with leading organizations such as Deutsche Borse, Culture Amp, eDreams ODIGEO, IHOP, IPG Mediabrands, HSBC, John Lewis Partnership, Macquarie Bank, Priceline, The Knot Worldwide, Shopify, the Singapore Government, U.S. Steel, & Wendy's. At Google Cloud Next '23, they announced new relationships with The Estée Lauder Companies, FOX Sports, GE Appliances, General Motors, HCA Healthcare, etc. CEO Thomas Kurian thanks every customer as they keep trusting in Google Cloud.

At Google Cloud Next ’23, they said that they are now glad to announce new ways which are beneficial for each business, government, and user benefit from generative AI & famous cloud technologies:

 

• AI-optimized Infrastructure:

  This most advanced infrastructure helps companies to train as well as serve models. The infrastructure is offered in the cloud regions to run in the data centers with Google Distributed Cloud.

 Vertex AI:

• Developer tools can be used for creating models & AI-powered applications along with major advancements in Vertex AI, to generate custom models and produce custom Search and Conversation apps using enterprise data.

 Duet AI:

• This AI collaborator is integrated deeply into Google Workspace & Google Cloud. It can provide users a project manager, a writing helper, a spreadsheet expert, a note taker for meetings, as well as a creative visual designer. This one can collaborate from different perspectives. For instance, it can be an expert coder, a database pro, a software reliability engineer, an expert data analyst, and a cybersecurity adviser.

 New infrastructure and tools to help customers:

Gen AI has become too revolutionary for the broad applications and advanced capabilities that require a sophisticated infrastructure. It has been 25 years since Google Cloud continues to invest in its data centers & networks. Currently, they have a huge network expanding almost 38 cloud regions. The aim of Google Cloud is to operate totally on carbon-free energy 24/7 by 2030.

 The AI-optimized infrastructure is an excellent choice that helps to train and serve gen AI models. In this case, it needs to be mentioned that more than 70% of gen AI unicorns are Google Cloud customers such as Cohere, AI21, Jasper, MosaicML, Replit, Runway, Anthropic, & Typeface. At Google Cloud Next '23 event, they announced crucial infrastructure advancements to assist customers, like:

 

• Cloud TPU v5e: It is the most versatile and affordable AI accelerator. As a customer, it is possible to use a single Cloud TPU platform with the intention of running large-scale AI training & inference. This one is optimised for efficiency and can scale to tens of thousands of chips. Regarding training performance for each dollar, this one can offer up to two times more improvement than a Cloud TPU v4.

 A3 VMs with NVIDIA H100 GPU:

• It is powered by NVIDIA's H100 GPU and is expected to be available in the next month. This one is made with high-performance networking for enabling the popular gen AI and LLM innovations. Thus, organizations can achieve training performance three times better than the prior generation A2.

 

• Cross-Cloud Network: This one is a global networking platform where people can connect applications & secure them across clouds. The platform can provide ML-powered security for offering zero-level trust. This network is designed in a way that allows customers to access Google services from any cloud. Besides, it helps to decrease the level of network latency by up to 35%.

 

• Google Distributed Cloud: The purpose of its design is to meet the organization's demands. By enhancing the GDC portfolio, Google Cloud is trying to bring artificial intelligence to the edge.

 Vertex AI platform gets even better:

Vertex AI, which is the most comprehensive AI platform, is their best infrastructure, allowing people to create, deploy and scale ML-based models. People can access over a hundred foundation models like third-party open-source versions. They are optimized for various tasks like text, chat, images, speech, software code, etc. Sec-PaLM 2 and other industry-specific models are offered for cybersecurity. Thus, it becomes possible to empower Tenable & other global security providers and life science companies such as HCA Healthcare, and Meditech.

 With the help of the Vertex AI Search & Conversation, organizations are able to generate Search & Chat apps within minutes with a little bit of coding, enterprise-grade management & security built in. Also, it is possible to get friendly tools from the Vertex AI Generative AI Studio to customize models with enterprise-grade controls. Developers can create sophisticated applications depending on semantic understanding of text or images and RLHF, using developer tools such as Text Embeddings API.

 

• Digital Watermarking on Vertex AI: It is powered by Google DeepMind SynthID, which can provide state-of-the-art technology to embed watermarks into images of pixels. As a result, it helps to make it invisible to the human eye and hard to tamper with. This watermarking can give you a scalable approach to recognising AI-generated images responsibly. Google Cloud is the first cloud provider to deliver the technology for AI-generated pictures.

 

• Colab Enterprise: It is a managed service combining enterprise-level security & compliance abilities and Google's Colab notebooks. Using this, service data scientists can accelerate AI workflows while having access to Vertex AI platform capabilities.

 The Vertex AI is designed to provide the users full control of  their data, code, and IP. If you customize and train a model with Vertex AI, the data will not be exposed to the foundation model. The model's snapshot enables you to train & encapsulate this together in a private configuration.

Duet AI in Google Cloud

 Hence, Duet AI in Google Cloud announcements come with some advancements for:

 

• Software development: It can offer expert assistance across the whole software development lifecycle. As a result, developers can be more productive by following the approach to minimize context switching. Customers can also modernize applications with the help of code refactoring. In addition, the Duet AI in Apigee allows developers to produce APIs and integrations.

 

• Application & infrastructure operations: With Duet AI, operators are able to chat in natural language across several services in the Google Cloud Console.

 

• Data Analytics: In BigQuery, Duet AI can help contextually write SQL queries & Python code. Besides, it can build full functions & code blocks, suggest code completions automatically, etc. Depending on the metadata & scheme, it produces recommendations. The capabilities let data teams focus more on the results.

 

• Accelerating & modernizing databases: Duet AI in AlloyDB, Cloud Spanner and Cloud SQL is useful for producing code with the intention of structuring, modifying, or querying data. Google Cloud has brought the Duet AI's power to DMS for automating the conversion of database code.

 

Conclusion:

 Currently, humans are in the digital transformation era, which is getting charged or fueled by gen AI.

Technology helps to improve the way businesses operate as well as the way humans interact with each other. In this event, it is said that Google Cloud is grateful for the opportunity to be on this journey with their customer.

 

Google Cloud Natural Language API

Google’s Cloud Natural Language API

Cloud Natural Language API, established by Google is said to provide customers with language analyser which according to the company `reveals the structure as well as the meaning of your text. The public beta launch of Cloud natural language API is a new service giving developers access to Google-powered emotion analysis, entity, recognition together with grammar analysis.

 Some of this tends to gauge believing some words positive and the others negative. When observed by Motherboard it was found that the analyser of Google interpreted some words like homosexual to be negative. This is evident that the API that tends to judge depending on the information it has been fed, now seems to give out partial analysis. The tool has been developed to provide companies with a preview on how their language would be expected.

Editing complete sentences would provide predictive analysis on each word and as the overall statement on a negative to positive scale, respectively. AI systems have been trained in utilisation of texts, media as well as books given to it.

Whatever Cloud Natural Language API consumed to form its criteria in assessing English text for sentiment, it influenced the study to negative attribution of certain descriptive terms. No confirmation has been provided by Google to Motherboard as to the body of the text fed to the Cloud Natural Language API.
 
API Connects Other Pre-Trained Machine Learning API
 
Once it begins to engage content from the outside world even if begins with a remote set of contents to comprehend sentiments, it tends to get polluted with the negative word connections found in it. A confirmation had been given by Google to Motherboard that its NLP API had produced biased results in a statement.

There had been clear parallels with the ill-fated as well as impressionable AI chatbots Tay of Microsoft, which had been rapidly, pulled offline by the company in March 2016 after the users of Twitters had taught it to be a shockingly racist as well as sexist conspiracy philosopher.

In July, Google had tried once again with its bot Zo which had learned the same horrible habits form human and had to be quickly shut down. The new API connects the other pre-trained machine-learning API of Google such as the Cloud Speech API which has been made available in public beta, together with the Vision API and the Translate API.
 
Assist Text in English/Spanish/Japanese
 
The latest Cloud Natural Language API presently tends to assist texts in English, Spanish and Japanese. The purpose of Google here is to provide a service which could meet the scale as well as the performance essential for developers and enterprises in a comprehensive range of industries. Providing API for sentiment analysis and entity recognition is not new where services like Thomson Reuters Open Calais have been providing assistance for entity recognition for around ten years now. Sentiment analysis is also not a new concept. On the other hand, grammar analysis API which tend to classify parts of speech and develop dependency analyse trees are not as extensively available still. It would be interesting to know how developers would be utilising these apps though it is easy to see how the same could be utilised to power chat bots for instance and support them in comprehending incoming request.

Google Antes Up Its Own Cloud Migration Appliance

Google is bringing its own data transfer appliance for cloud migration

When we talk about the cloud migration hen the toughest challenge to overcome is to ensure reliable and consistent migration methodology. Moving databases and data centers are not an easy game even for the proficient administrators. This is faced by almost all the major companies and new start-ups when they are trying to build new application or make use of new data residing in the cloud. Data migration between two points is always seen as tough egg to break due higher costs and huge time consumption this is where cloud vendors come into the play.

Remedy for data to cloud migration problem

The problem faced in the cloud migration is quite incomprehensible even with modern technologies at disposable. If a person has 10 Gbps connection then transferring petabyte of data from any data center will consume as many as 12 days to put it on the cloud. In the old golden days companies used to Sneakernet in the sky wherein a pile of data is loaded on a secure disk and it is shipped off to any of their cloud vendor. Microsoft Azure made use of this system for quite some time before a new system came into being called Snowball and this method is also used by the Amazon AWS. This doesn’t mean that sending secure disks to the cloud vendor has become obsolete.

Google is moving into the enterprise cloud migration business

The tech giant is getting into the enterprise cloud business on a serious note but quite specifically in the migration appliance space. In this end of the business the consumer arc is very similar to online shopping portal. User goes online, orders the device and this particular device is made available for definite set of time after which they have to send it back to the provider.

Cloud migration market is buzzing with immense order in the one petabyte category therefore Google is introducing two models with size of 100 and 480TBytes. On other hand its competitors Amazon tends to work in lower end of the spectrum with 3 models with data size of 50, 80 and 100 TBytes units. Amazon is way ahead in the field of data transfer and it has developed a number of solutions for high end data migration with Amazon Snowmobile wherein it brings a 12 wheeler 45-foot container for transferring 100 PBytes of data.

Pricing will be a key to grab a piece of the cloud migration market

Google is mainly focusing on cornering market for the petabyte cloud migration. Therefore it has brought on par pricing with Snowball for the smaller 100 TByte units and in order to make migration appliance more appealing to the users by keeping it 35% below the rival offerings. When it comes to design Google is going for plug-in based form and function while Amazon brings self-standing units.

However Google has revealed many details of its devices, service offerings, capabilities and benefits but it seems like it is eagerly looking forward to give stiff competition to market leaders namely Amazon and Microsoft Azure.