They ‘‘Treat Google like a mountain. You can climb the mountain, but you can’t move it’’, these are the wise words from Jeff Bezos, the CEO and founder of Amazon. It’s yet again proving to be true in the case of generative AI solutions.
When OpenAI came out with ChatGPT, an AI chatbot based on a large language model, it kind of removed Google from the equation for a short while. Well, Google is settling the score straight with its new announcement, Gemini – Google’s most advanced LLM-based AI solution. There is a lot to say about this new addition by Google and this blog will give you the flashback of A cold war between tech giants, how Google came up with Gemini and what makes Gemini special.
The Epic Ai Chatbot Race: Story About Google, Meta And Other Beasts In The Ring
It took the world by surprise when a start-up which got its origin from a non-profit organisation was able to come out with one of the best AI chatbot solutions. Adding to the amusement, ChatGPT was far superior in terms of capabilities when compared with the public versions of AI chatbots from tech titans like META and Google. For instance, META’s BlenderBOT was a debacle in terms of reasoning, natural response and avoiding discrimination. Even though they came out with other variants of chatbots, soon they all were pulled off the grid due to issues in the quality of response. Currently blenderbots prototype is only available in the US.
ChatGPT was an overnight success and the tech world took notice of the same. According to OpenAI, ChatGPT acquired 1 million users just 5 days after launching in November 2022. By comparison, it took Instagram approximately 2.5 months to reach 1 million downloads. So by all metrics, this is an instant hit. According to sources closer to OpenAI, even the management of OpenAI was surprised to find such a massive success.
Soon after the public release and success of ChatGPT 3.5, google tried to put together all its backstage works on an AI chatbot and introduced Bard based on LaMDA. Well, the public soon realised it was half-cooked. The most awkward of all the wrong answers Bard gave was the answer to the question “What new discoveries from the James Webb Space Telescope (JWST) can I tell my 9-year old about?” – because it was published in the announcement tweet another promotion materials by google! This simple error resulted in an eye-watering $100 billion drop in Alphabet’s share value.
Google learned the lesson, they kept Bard running but continued optimising their best AI solution Gemini in the background. In the Google I/O 2023, Sundar Pichai teased about Gemini and on the 6th of December 2023 it was officially announced in all its glory. We will get to know more about this model in the upcoming months but as of now, we are informed that it’s behind the new and improved Bard. Soon we will see more applications of Gemini for research, enterprise and individual requirements.
The Story Behind Gemini
Google Gemini is not something which came out just because of ChatGPT or any other potential AI solution, it was already in the works and Google waited till it was mature enough to release. They spend a significant amount of time and resources on developing Gemini.
Google had a strong team of researchers and scientists working on various AI projects and the key companies responsible for AI & ML development were Google Brain and Google DeepMind. On April 2023, 5 months after the release of ChatGPT 3.5 research preview, Google merged Google Brain and Google DeepMind. It was a major move to bring together all the capabilities under Google roofs to bring out a stable and superior AI solution. According to Demis Hassabis, CEO and Co-Founder of Google DeepMind, the Gemini team received support from multiple verticals of Google including Google Research.
The dedication and hard work of many devoted programmers, researchers and scientists gave birth to Gemini and its multiple variants. From the initial testing, we found the Gemini-powered Bard is better in terms of reasoning, natural response and coherence than the previous version powered by LaMDA.
What Makes Gemini Special?
Google Gemini is different from all the other models on several grounds. We will go through them briefly.
Gemini is multimodal by design: Compared to other models which have multiple models wielded together for understanding and responding to various inputs, like text, image and video, Gemini has the advantage of being the model which is natively multimodal. Gemini can accept, interpret and respond to multiple inputs by design. This is significant because when Gemini is scaled down to its nano variant, it can function independently for various purposes other than just text output or audio response.
Most flexible model by Google: Google Gemini comes in 3 variants as of now. Gemini Ultra for enterprise-level utilities, Gemini Pro for business-level and professional applications and Gemini Nano for applications in mobile and other consumer technologies. Gemini is the first of its kind to be introduced in multiple variants. There are indeed models which got enterprise versions and public versions, but Google pushes the benchmark further with Gemini Nano, a variant exclusively crafted for the large mobile market. In terms of business, it is a wise move!
iii. Performance superiority: As of now, we only got the test results from benchmarking tests done by Google engineers, if we go by their word, out of all 32 industry-leading academic benchmark tests, Google Gemini came out to be the best-performing model in 30 of them. Google Gemini Ulta is the first AI to outperform human experts on MMLU (massive multitask language understanding), which uses a combination of 57 subjects such as math, physics, history, law, medicine and ethics for testing both world knowledge and problem-solving abilities. Google Gemini Ultra scored 90% in MMLU whereas GPT 4 stands at 86.4% and a human expert is at 89.8%.
Powerful coding capabilities: Gemini can understand, explain, and generate code in popular programming languages like Python, Java, C++, and Go. Microsoft utilized the abilities of GPT 4 by developing GitHub copilot, which is now a go-to tool for many developers while coding. Google can utilise the abilities of Gemini to develop a similar product or even a better version to help developers solve coding hiccups and develop projects in the most optimized manner.
Integration with Google products: The biggest pro for Gemini is that it’s from the house of the world’s biggest tech conglomerate, and Gemini’s flexibility makes it easy for different product development teams at Google to multiple tech products ranging from email services to search to mobile devices. Google is already working on integrating Gemini with various Google products, such as Smart Reply in Gboard. This will allow users to take advantage of Gemini’s capabilities in their everyday lives and help google to improve Gemini with enormous user data and feedback in real-time.
Here it is, the most summarized article on the past, present and future of Google Gemini. We will follow up with more insightful articles on AI and other cutting edge, so don’t forget to follow us on socials to know about our next blog post.
If you are interested in AI, ML and Data Science, do visit our website to learn more about our courses specifically tailored for working professionals and learners who are looking for AI, ML and Data Science University level qualifications.
These programs are designed for working professionals and you can complete the program while working full time, thanks to our state-of-the-art online learning management system. Courses such as Masters in AI in Business are crafted for professionals with no coding background. If you are interested in knowing more, please fill up with form given below or kindly connect with us via WhatsApp.