Google Unveils Gemini: AI Tech Possibly Surpassing OpenAI's GPT-4
Austin JayThis week, Google introduced "Gemini," a formidable competitor to OpenAI's GPT-4.
Gemini comprises three distinct models varying in size and capabilities. As per Google's announcement, its most advanced version, Gemini Ultra, is geared for intricate tasks and surpasses GPT-4 in various domains, such as historical knowledge, legal expertise, Python code generation, and multi-step reasoning.
Google highlighted Gemini's outperformance of GPT-4 on the Massive Multitask Language Understanding (MMLU) test, a widely recognized benchmark assessing AI models' knowledge and problem-solving abilities.
Gemini Ultra, although not yet accessible to the public, promises significant advancements in AI capabilities.
Kevin Roose said in a podcast on Hard Fork that he likened the Massive Multitask Language Understanding (MMLU) to the "SATs for AI models."
The MMLU encompasses 57 subjects, spanning math, physics, history, law, medicine, and ethics, assessing broad knowledge and problem-solving skills.
Introducing Gemini, Google’s largest and most capable AI model. 🧵 #GeminiAI https://t.co/T0tIw9HQyO
— Google (@Google) December 6, 2023
Gemini Ultra achieved a 90% score on the MMLU, surpassing GPT -4's 86.4%, marking an essential milestone as the first model to outperform human experts, who scored approximately 89.8%.
This achievement led Roose to speculate on artificial general intelligence (AGI).
While GPT-4 excelled in everyday common sense reasoning, Gemini Ultra was praised for its natively multimodal design, enabling seamless processing of various data types-text, audio, code, images, and video-without the need for combination from separate models.
Google claims Gemini's built-in multimodality offers superior input comprehension compared to other multimodal models.
Additionally, Google suggests that Gemini's architecture and computing power could outperform GPT-4, a point echoed by analysts at SemiAnalysis.
Also Read: Google Drive Users Experience File Loss, Increasing Concerns Over Disappearing Data
The generative AI market is projected to explode, reaching $1.3 trillion by 2032 from $40 billion in 2022.
One source exclaimed their forecasts that generative AI software will generate $280 billion in revenue by 2032. It is pivotal to Google's Gemini models, fueled by its own Tensor Processing Units.
Gemini Pro launches on Google Cloud, followed by Gemini Ultra after safety checks. These models boost Google's cloud business and refine applications like Bard.
Integrating Gemini into products like Search, Ads, and Chrome is strategic in facing AI's impact on search. Gemini Nano aids Android 14 smartphones, enhancing on-device AI features in Pixel 8 Pro devices.
Using custom AI chips over Nvidia GPUs, Google gains a cost advantage amid the AI surge, which is crucial in training advanced generative AI models.
In AI, Google's Gemini shows up as a model that can handle text, images, and sound simultaneously, giving it an advantage over GPT-4, which works with words. It says it can solve problems like a human, which makes people think AI can deal with complicated stuff no one thought it could do before.
But is Gemini really going to do what it claims? Is it better than GPT-4? There are still concerns about powerful AI and what could go wrong.
The start of Gemini means a big moment for AI stuff. Folks are curious about what kind of effect it might have. The writer invites people's thoughts on how Gemini could change things and wants discussions on where AI is headed.
As the journey for AI keeps going, the questions stay open without answers for now until later on down the road. The article ends by asking readers to share and support it, realizing that getting people involved helps spread the word further.
Related Article: Google Launches Gemini: GPT-4 Competitor Available Now
most read
related stories
more stories from News
Discover the latest Google Pixel update! Get the scoop on security, stability, and build numbers. Don't miss out, read now to know how you can elevate your Pixel experience.
ernest hamiltonKnow the latest on TikTok's battle against misogyny and harassment as Tobias Henning leaves the company. Read the story now.
ernest hamiltonStay updated! Google Messages Beta introduces a fresh Gallery Picker UI. Explore it now for enhanced messaging experience.
ernest hamiltonExciting news! Samsung has begun developing One UI 7.0 Beta, based on Android 15, for the Galaxy S24 series. Stay tuned for the latest updates.
ernest hamiltonStay tuned for Pixel 8a's first update, introducing Google's AI Wallpaper Generator! Don't miss out on this exciting feature.
ernest hamiltonWith Google Messages, you can now enjoy a new audio recorder and voice moods, enhancing your messaging experience. Check it out today!
ernest hamiltonThe iOS 17.5 RC release brings several enhancements, including updates to the Podcasts widget, an improved unwanted tracking system, Repair State mode, and more.
ernest hamiltonThe decision to revoke export licenses to Huawei could have significant implications for chip sales by Intel and Qualcomm in the US.
ernest hamilton