Google has launched its Gemini AI. Is it a ChatGPT killer?

Google has thrown down the gauntlet to OpenAI. On Wednesday, the tech giant unveiled an AI called Gemini – a rival to ChatGPT which rocketed to global fame last year. “We’re approaching this work boldly and responsibly,” Google CEO Sundar Pichai wrote in a blog post. “That means being ambitious in our research and pursuing the capabilities that will bring enormous benefits to people and society while building in safeguards and working collaboratively with governments and experts to address risks as AI becomes more capable.” “This is a significant milestone in the development of AI, and the start of a new era for us at Google,” said Demis Hassabis, CEO of Google DeepMind, the AI division behind Gemini. “I am in awe of what it’s capable of,” Google DeepMind vice president of product Eli Collins said of Gemini. But is Gemini better than ChatGPT? Let’s take a closer look: First, let’s briefly examine Gemini. Collins claimed Gemini is the first AI model to outperform human experts in certain benchmarks involving problem-solving, math, physics, history, law, medicine and ethics. A demonstration showed Gemini recognising what it was shown, from a person acting out a “Matrix” movie scene to someone drawing a duck and then holding up a rubber duck.

Gemini commented on what it was shown, making comparisons, drawing conclusions, and offering suggestions.

Google will roll out the Gemin in phases – the ‘Nano’ and ‘Pro’ versions will immediately be integrated into Google’s AI-powered chatbot Bard and its Pixel 8 Pro smartphone. Then, next year, Gemini’s ‘Ultra’ model will launch ‘Bard Advanced’ – a juiced-up version of the chatbot that initially will only be offered to a test audience. The AI, at first, will only work in English throughout the world, although Google executives assured reporters during a briefing that the technology will have no problem eventually diversifying into other languages. [caption id=“attachment_13475452” align=“alignnone” width=“640”] Google on Wednesday launched Gemini, an AI chatbot that they claim is better than ChatGPT.[/caption] Gemini will also eventually be infused into Google’s dominant search engine, although the timing of that transition hasn’t been spelled out yet. Google raced out its own Bard chatbot earlier this year, continually updating the chatbot based on people’s feedback, according to Bard vice president Sissie Hsiao. “All of that rapid innovation is bringing us to what we see as a truly transformative moment,” Hsiao said during the briefing. “With Gemini, Bard is getting its biggest upgrade yet.” Bard will use Gemini for more advanced reasoning, planning, and understanding capabilities, a demonstration showed. Gemini-infused Bard will be expanded to be “multi-modal,” meaning it will be able to work with auditory and visual input as well as text prompts, executives said. “With Gemini we are one step closer to our vision of bringing you the best AI collaborator in the world,” Hsiao said. Gemini ramps up the quality of Bard’s performance, whether in writing poetry or computer code to shopping queries or research projects, according to Hsiao. Is it better than ChatGPT? The benchmarks would certainly indicate so.

Google Gemini's benchmark numbers absolutely CRUSH GPT-4!!!!

We have a war on our hands. pic.twitter.com/AJ1meXVqSq

— Deedy (@deedydas) December 6, 2023

While Google declined to share Gemini’s parameter count — one but not the only measure of a model’s complexity – Collins claimed performance of the ‘Ultra’ version of Gemini ‘far exceeds’ that of other state-of-the-art models in 30 benchmark tests measuring capabilities such as image understanding or mathematical reasoning. But there’s a caveat.

A piece in tech.co pointed out that Google said that only the ‘Ultra’ version of Gemini can beat ChatGPT.

More from Explainers

FirstUp: South Korea’s Lee Jae Myung to meet Trump in Washington and other top news Sitting all day in your office AC? Why prolonged exposure can make you sick

Indeed, a white paper released Wednesday outlined the most capable version of Gemini outperforming GPT-4 on multiple-choice exams, grade-school math and other benchmarks. However, it acknowledged ongoing struggles in getting AI models to achieve higher-level reasoning skills. Meanwhile, the old problems continue to rear their heads. Nilay Patel in _The Verg_e wrote that he “asked the new Gemini-powered Google Bard an important question and it confidently hallucinated a Vergecast interview that doesn’t exist complete with a link.” In The Atlantic, Matteo Wong wrote that while Gemini can beat GPT-4 ‘on most parameters’ it is just an “iterative advance”. Translation? Gemini isn’t one giant leap for mankind – yet. [caption id=“attachment_13452662” align=“alignnone” width=“640”] Experts say companies are in a take-no-prisoners arms race.[/caption] Besides, OpenAI has an ace up its sleeve. “We’re in a kind of tit-for-tat arms race,” Oren Etzioni, a professor emeritus at the University of Washington and former CEO of the Allen Institute for AI told Wired. “There’s no reason to disbelieve that Gemini does better than GPT-4 on these benchmarks, but the next version, GPT-5, will do better than that.” “This is a take-no-prisoners, must-win war,” Etzioni added. ‘We don’t really know what’s inside’ But some computer scientists see limits in how much can be done with large language models, which work by repeatedly predicting the next word in a sentence and are prone to making up errors known as hallucinations. “We made a ton of progress in what’s called factuality with Gemini. So Gemini is our best model in that regard. But it’s still, I would say, an unsolved research problem,” Collins said. The staying power of generative AI chatbots, once the initial excitement has faded, is yet to be confirmed. Moreover, integration of the OpenAI-based chatbot into Microsoft’s search engine earlier this year failed to make an impact on Google’s overwhelming dominance of search. Governments and tech companies, however, insist that generative AI is technology’s next big chapter and have ramped up spending on new products, research, and infrastructure. “That’s the problem with all these proprietary models,” Alexei Efros, a professor at UC Berkeley told Wired. “We don’t really know what’s inside.” With inputs from agencies