Google has recently launched a new AI model currently named ‘Gemini’ and is planning on taking first place. It is said to outperform both OpenAI’s GPT- 4 model and even its “expert level” humans after a range of intelligence tests were conducted.
CEO Sundar Pichai revealed this information during a Google conference back in May of this year, but that was when it was in the early stages of development and now it has been confirmed that Gemini is ready to launch its new competitive model to the public soon.
What Are The Features of Gemini?
There are currently three versions of Gemini that have specifically been developed for different applications, called Nano, Pro and Ultra, which all increase in size and capabilities. Google has so far refused to go into details on the size of Pro and Ultra, the number of parameters, the scale or the source of their training data.
Its smallest version, Nano, is designed specifically for mobile phones and is available in two models: one for slower phones with 1.8 billion parameters and one for modern devices that have 3.25 billion parameters.
Parameters are anything; for example, a file name, a range of values, a code—it is essentially anything you can think of. A function can carry out operations using parameters even if it is not aware of the precise input values in advance. Programmers use functions to separate their code into logical blocks, and parameters are essential parts of those functions.
Why is Gemini Better?
When comparing Gemini to its rival ChatGPT, their AI models are an inexact science but when you look at the numbers, there is a significant difference. GPT-4 is rumoured to include up to 1.7 trillion. That’s just a comparison to its Nano and not its other models.
The mid-range Pro version beats other models, but the Ultra exceeds them all. Google claims that it scored 90 per cent on the industry-standard MMLU benchmark, whereas an “expert level” human is expected to achieve 89.8 per cent. As a small victory but still a victory, this has made history as the first AI to beat a human at the test. In the same score, GPT-4 scored 87 per cent and LLAMA-2 scored 68 per cent.
All current AI systems are just scratching the surface of multimodal technology; their content and data formats are not efficient. The human brain is complex and can interpret and understand various data formats, including text, words, sounds and visuals. We can make sense of the world around us, respond to stimuli, solve problems in unique ways and spot online scams—that is what Google has achieved with Gemini.
Conclusions
Google’s groundbreaking AI, Gemini, surpasses predecessors like GPT-4 and even outperforms humans in intelligence tests. Its diverse versions boast impressive capabilities, but amid this progress, there’s a need for caution. As AI evolves, there’s a potential for misuse in online scams, including forex scams. While Gemini signifies a milestone in AI, vigilance, education, and strict regulations are essential to prevent exploitation and scams that might arise from its advanced capabilities. Balancing progress with caution is crucial in navigating the impact of emerging AI like Gemini.