The specialized agents, Gemini 1.5 Pro and Code Assist. Here is Google’s strategy for bringing AI into the company

LAS VEGAS – Gemini is now everywhere. The mission here in Las Vegas was to demonstrate to the market that inside Google’s “cloud” there is already a clear strategy on how to bring generative artificial intelligence into the company. At the Mandalay Convention Center in front of thousands of partners and companies, the leaders of the Mountain View giant introduced the work of Google Cloud Next with a flurry of announcements to convince the market and customers of the progress of the integration of Ai gen to the inside each of their services. Because now more than ever for AI providers, as Google Cloud number one Thomas Kurina specified, the real challenge is to put the business applications of generative artificial intelligence on the ground (on the cloud): “More than 60% of startups and almost 90% of unicorns (statuses that have exceeded one billion) are Google Cloud customers including companies like Anthropic, AI21 Labs or the French Mistal are training their algorithms in our infrastructures”. Google Cloud Next this year is a two-day event of use cases and business applications together with the names Bayer, Cintas, Discover Financial, IHG Hotels & Resorts, Mercedes Benz, Palo Alto Networks, Verizon, WPP also to send a message to the clouds competitors, first of all Microsoft’s Azure and AWS. Gemini, Google’s large language model (LLM) has so far not impressed. As Sundar Pichai also admitted to Google, the success of ChatGpt surprised the Group which has been working on AI for several years and forced them to speed up by perhaps launching products like Bard too hastily. Things went better with Gemini, at least from a technological point of view, but the problem seems to be the organizational structure of Mountain View with too many leaders working independently and in competition and the fear that by focusing on Ai we might question or even cannibalize the their business model based on advertising and search engine. Google Cloud Next appears to be a muscular response from its number one in person who from June onwards opened the drawers of its research laboratories to integrate everything that its chatbots can offer into its offer for companies. Let’s see what’s new.

What are specialized agents.

They are specialized chatbots that can be “trained” and managed by companies. There are three types: Customer agent, Employee Agent e Creative Agent. The first are assistants like those being created for the travel industry who guide you in choosing your holiday. They can help customers interact with companies more seamlessly by working across multiple channels (web, mobile, call center and point of sale) and in multiple modalities, such as text, voice and more. The Employee Agent they help employees be more productive and work better together as in the case of Etsy which uses platform training for businesses Vertex Ai, the heart of the enterprise offering, to optimize search and ad recommendation models, providing buyers with suggestions to improve listings and increase sales. And then there are the Creative Agent. During a presentation, their functioning was shown. The agent studies images and presentations of the company and on the basis of this information proposes advertising or marketing campaigns that take inspiration from what the company has done so far. The demo shown is surprising because you start with an idea and in a few seconds you have something visual to communicate.

Gemini 1.5 Pro is among us.

The most awaited innovation was the debut of Gemini 1.5 Pro. It is presented as Google’s first larger model and completely multimodal, meaning it allows you to manage text, images and, for the first time, videos in input. Customers have the ability to manage up to one million tokens, which translates into the ability to work with texts of over 700,000 words, 11 hours of audio and 1 hour of video. It should be remembered that, at the moment, ChatGPT web version reaches 8,000, Gemini 1.0 Pro at 32,000, GPT-4 Turbo at 128,000 and Claude 2 at 200,000.

Google lancia Image 2

Let’s say that following the controversy over the excessively inclusive Nazi soldiers and the consequent stop, image generation is not Google’s strong point which was forced to take some steps backwards with Gemini. In anticipation of its return, the company is launching an advanced image generation tool, Imagen 2, within its Vertex AI development platform, although it’s a decidedly more enterprise-oriented tool. Of interest to business types, Imagen 2 can reproduce text, emblems, and logos in multiple languages, optionally overlaying such elements into existing images, for example, on business cards, apparel, and products.

At the Vertex Ai center, the news

The enterprise platform is, as expected, the true protagonist of the event. Google Cloud offers more than 130 proprietary and third-party models and the idea is to make it the real competitor to Azure and AWS. There are many announcements in this sense. Like the landing of Claude 3, the Anthropic model, currently not available in Italy, for customers on Vertex AI. As well as CodeGema, lighter and open source systems for generating code for programmers. The project is to equip the platform and therefore companies with tools to generate their own chatbots internally, ensuring privacy and security. There has been no mention of search engines for now. The Financial Times had anticipated a project to make search powered by paid AI. But within Vertex numerous innovations have been presented that connect company data and Google’s advanced search services. As for competitors at this stage, the challenge is to connect company data with the potential of AI and information search functions. On this frontier, Google wants to play its cards by strengthening the Vertex platform.