Google I/O 2024: Google’s series of AI announcements, Gemini Flash, AI search

Google I/O 2024: Alphabet, Google's parent company, on Tuesday, showcased its advancements in artificial intelligence across various business sectors with updates including improvements to the Gemini chatbot and its search engine, as part of its efforts to stay competitive in the AI space.

Google's announcements come in response to the success of OpenAI’s ChatGPT, which launched in 2022 with backing from Microsoft, posing a challenge to Google’s dominance in online search and AI. Key among Google’s new developments is a faster, more cost-effective version of its Gemini 1.5 AI models, known as Flash. Additionally, a prototype named Project Astra can engage with users in real-time based on images from their smartphone cameras, and search results will now feature AI-generated headlines.

Alphabet CEO Sundar Pichai described this phase as a "moment of growth and opportunity" when questioned about the potential impact of AI updates on Google's profitable business.

The announcements were made at Google's annual I/O developer event in Mountain View, California, a day after OpenAI's shorter showcase. OpenAI demonstrated how ChatGPT can provide spoken answers with human-like intonation to various prompts. OpenAI CEO Sam Altman mentioned that their software feels like "AI from the movies."

Competition between AI giants

Google's presentations often paralleled those of OpenAI, highlighting the intense competition between the two companies. For instance, Google DeepMind has been developing technology for daily consumer tasks, as seen with Project Astra, which uses smartphone cameras to interpret surroundings. A demo showed how it identified a speaker and located misplaced glasses. 

Google also hinted at integrating Project Astra with Gemini Live, a more natural-sounding voice and text assistant than previous versions of Google Assistant. Demis Hassabis, CEO of Google DeepMind, explained that Project Astra aims to be a "universal AI agent that can be truly helpful in everyday life."

Veo AI video generator 

In video generation, Google introduced Veo, an AI model capable of producing 1080p-resolution videos over a minute long, available to select creators like filmmaker Donald Glover. This is in direct competition with OpenAI’s similar software Sora, which has captured the attention of Hollywood.

“It accurately captures the nuance and tone of a prompt, and provides an unprecedented level of creative control — understanding prompts for all kinds of cinematic effects, like time lapses or aerial shots of a landscape,” said Google. 

Gemini gets better with Gemini Pro 1.5

Google also improved its Gemini Pro 1.5 model, now capable of handling twice as much data, up to 2 million tokens. The improvement allows the AI to process extensive text or lengthy videos. The Pro model, initially supporting prompt sizes up to 1 million tokens, will be accessible to subscribers of Google’s Gemini Advanced service. Following these announcements, Alphabet’s shares rose by 1 per cent to $172.59 on Tuesday afternoon.

Google has also introduced Gemma, a family of lightweight open models based on the same research as the Gemini models. 

New computing chips to power search engine

Google revealed efforts to improve AI with new computing chips and to revamp its search engine. The sixth-generation tensor processing unit (TPU) aims to provide an alternative to Nvidia's dominant processors and will be available to Google Cloud customers by late 2024.

For US users of Google Search, the company will soon employ AI to organise search results for queries related to dining, recipes, and eventually movies, books, and other content. AI Overviews, which synthesise information to answer complex queries, will be available to all US users this week, following extensive public testing since last year’s I/O event.

Jacob Bourne, an analyst at eMarketer, noted that the reception of AI Overviews will indicate Google's ability to adapt its Search product to the generative AI era. He added that to maintain its competitive edge and satisfy investors, Google must focus on translating AI innovations into profitable products and services at scale.

Ads to continue 

Advertisements will continue to appear throughout Google’s web pages, and AI Overviews will be rolled out to over a billion users by the end of the year. In 2023, Alphabet reported revenue of $307.4 billion, with the majority coming from ads on Google Search and other properties.

Additionally, Google is testing a feature allowing users to ask questions about videos they upload to Google Search, similar to existing image-based queries. A demonstration showed how this could help diagnose issues with a broken record player.

(With Reuters inputs)