By Kawtar Nasser | Staff Writer
You may have heard about Google Gemini but do you really know what it is?
What is Gemini AI?
Gemini AI is an advanced artificial intelligence system that enables multiple agents to engage in dynamic conversations. It has launched three models: Gemini Ultra as a leading model, Gemini Pro a “lite” version, and Gemini Nano which is a smaller model that runs on mobile devices. All three versions are multimodal which means that they were trained to work with and use not only words but also a variety of audios, images, and videos in addition to codebases and text in different languages. This family of new AI models from Google aims to emphasize features such as multi-agent interaction and dynamic personality simulation which makes interactions more human-like.
Using a transformative architecture, the AI will be able to understand things in a more intuitive way and not which is one of the drawbacks of using AI. This will give Google Gemini an edge; however, it still needs more attempts to play catchup as GPT models are already dominating the scene.
What are the features?
Gemini is set to be run on any device from data centers to phones. Gemini AI presents three distinct models catering to varied needs:
- Gemini Ultra, optimized for complex tasks, outperforms GPT-4 in language and multimodal benchmarks.
- Gemini Pro strikes a balance between scalability and performance, with a version used by Google Bard for handling intricate queries.
- Gemini Nano, designed for local operation on smartphones, enhances efficiency by executing tasks directly on the device, currently powering features like smart replies in Gboard on the Google Pixel 8 Pro.
Google’s Gemini vs OpenAI’s GPT-4
Technically speaking, Gemini has outperformed GPT-4 in 30 out of 32 benchmarks by having a clear advantage when it comes to dealing with images and videos in its latest model Gemini Ultra. However, Google acknowledges the competitive landscape with models such as GPT-4 from OpenAI which have gained major traction. This has created a sense of urgency and declared a “code red” after the launch of ChatGPT setting the race for AI competitive space.
Both Sundar Pichai, CEO of Google, and Demis Hassabis, CEO of Google DeepMind, emphasize a cautious approach, asserting that they are unwilling to rush developments merely to stay abreast of competitors, particularly as the prospect of achieving artificial general intelligence (AGI) draws nearer. AGI, characterized by self-improvement and surpassing human intelligence, holds the potential to revolutionize the world. Hassabis underscores the dynamic nature of AGI as an “active technology,” advocating for a balanced approach marked by caution and optimism as they navigate this transformative frontier.
Latest controversies
Google CEO Sundar Pichai has addressed concerns regarding biases and unacceptable responses from the company’s Gemini artificial intelligence model, which produced historically inaccurate portrayals, such as German World War II soldiers depicted as people of color. In a memo to employees, Pichai acknowledged the offensive nature of these outputs and emphasized the need for immediate action. The company has temporarily paused Gemini’s ability to generate images of people and is working to rectify the issues. This incident highlights ongoing challenges with AI systems reproducing biases present in their training data. Pichai outlined plans for structural changes, updated product guidelines, and improved launch processes, along with enhanced red-teaming to simulate potential misuse of the product. Despite efforts to mitigate bias, Google’s Gemini model reflects broader industry challenges in addressing these issues effectively.
Latest updates
Google’s announcement of Gemini Business, a new tier of its AI-capable enterprise productivity suite Gemini for Workspace, reflects the company’s ongoing competition with Microsoft in the generative AI era. Priced at $20 per user per month, Gemini for Business offers email drafting assistance, data analysis, and document creation, leveraging Google’s Gemini Ultra large language model. With no minimum user requirement and assurances of privacy, it directly competes with Microsoft’s Copilot for Microsoft 365. This move underscores the growing trend of companies integrating generative AI capabilities into their business products, reflecting investor expectations for tech firms to demonstrate returns on their AI investments.
References
(n.d.). YouTube: Home. Retrieved March 1, 2024, from https://www.forbes.com/sites/markminevich/2023/12/20/googles-gemini-challenging-openai-chatgpt-and-changing-the-game/?sh=6c5c643c62fa
Allyn, B. (2024, February 28). Google CEO Pichai says Gemini’s AI image results “offended our users”. NPR. Retrieved March 1, 2024, from https://www.npr.org/2024/02/28/1234532775/google-gemini-offended-users-images-race
Google Gemini: Everything you need to know about the new generative AI platform. (2024, February 16). TechCrunch. Retrieved March 1, 2024, from https://techcrunch.com/2024/02/16/what-is-google-gemini-ai/
Guinness, H. (2024, January 9). What is Google Gemini? What you need to know. Zapier. Retrieved March 1, 2024, from https://zapier.com/blog/google-gemini/
Milmo, D., & Hern, A. (2024, February 28). Google chief admits ‘biased’ AI tool’s photo diversity offended users. The Guardian. Retrieved March 1, 2024, from https://www.theguardian.com/technology/2024/feb/28/google-chief-ai-tools-photo-diversity-offended-users
Pierce, D. (2023, December 6). Google launches Gemini, the AI model it hopes will take down GPT-4. The Verge. Retrieved March 1, 2024, from https://www.theverge.com/2023/12/6/23990466/google-gemini-llm-ai-model