November 30th 2022 served as a catalysing moment for Big Tech: It was the date that OpenAI officially launched its artificial intelligence-powered chatbot, ChatGPT.

Amid an AI frenzy, Google announced its own competing generative AI system shortly after, which was also capable of understanding and responding to text-based prompts and queries.

The product in question, Bard—now known as Gemini—began its initial roll out to users on March 21st 2023. And with today being March 21st 2024, it’s now been an entire year since the product first became publicly available.

To commemorate Gemini’s one year anniversary, we’ve recapped key milestones thus far, heard from seasoned chatbot users about what they think of the product, and also posed the question: has the chatbot’s first year been a success?

But first, here’s a quick recap of Gemini and the technology underpinning it.

What is Gemini?

Gemini is a generative artificial intelligence chatbot. It can respond and reply to questions and requests on a diverse range of subjects, ranging from “what were yesterday’s top global news stories,” to “create a marketing plan for a new business,” to “explain quantum mechanics.”

Here’s an example in which, after our text-based prompt (prompts can also be sent via microphone, or accompanied by a photo—meaning it’s multimodal), Gemini explains what our website is after leveraging information from Google search.

The Gemini chatbot is ultimately an interface to a large language model—the technology that provides it with the ability to do and act. Large language models (LLMs) are trained on massive amounts of data, and are built on machine learning neural networks.

The model which powers the Gemini chatbot is also called Gemini, perhaps confusingly. As it stands, the latest version of the model is Gemini 1.5, a version that Google has described as “next-generation,” not least due to its mixture-of-experts (MoE) architecture. Demis Hassabis, the CEO of Google DeepMind—the company’s AI research lab—explains what this is here.

The Gemini model also comes in three variations: Gemini Nano, a smaller model for mobile; Gemini Pro, an all-purpose model for a wide range of tasks; and Gemini Ultra, the largest and most capable model of the three variations, built for more complex tasks.

All in all, Gemini is multifaceted: It’s a chatbot, a large language model, and a family of model variations, simultaneously.

However, it’s only been since February 2024 that the Gemini chatbot has been called Gemini, with Bard being its prior name. (More on this later.)

Key Milestones in Gemini’s First Year

Gemini’s first year is perhaps best described as mixed: While it’s been able to assist people across the globe in myriad impressive ways, there have also been some blunders, notably with some of the chatbot’s marketing and its image-generation capabilities.

Let’s take a look at these key milestones, both positive and negative, though starting with a story that technically took place before Gemini became publicly available—but was memorable nonetheless.

The James Webb Space Telescope Incident
Just days after Bard was announced in early February 2023, Google showed a small demo of Bard’s answering capabilities. However, the chatbot mistakenly presented inaccurate information regarding the James Webb Space Telescope.

Bard is an experimental conversational AI service, powered by LaMDA. Built using our large language models and drawing on information from the web, it’s a launchpad for curiosity and can help simplify complex topics → https://t.co/fSp531xKy3 pic.twitter.com/JecHXVmt8l

— Google (@Google) February 6, 2023

It wrongly asserted that the telescope took the very first photos of a planet outside our own solar system. It was, instead, the European Southern Observatory’s Very Large Telescope.

This then caused Google’s parent company, Alphabet, to lose around $100 billion in market value.

Bard’s Official Rollout
Bard began rolling out to the public, albeit in a limited way, on March 21st 2023—just over a month after it was first announced by Google.

While Google noted that Bard can be used to “boost your productivity, accelerate your ideas and fuel your curiosity” in its launch post, the company also took time to acknowledge that LLMs can face a range of issues, not least with inaccuracies and bias.

As was written last year: “Because they learn from a wide range of information that reflects real-world biases and stereotypes, those sometimes show up in their outputs. And they can provide inaccurate, misleading or false information while presenting it confidently.”

Bard initially utilised Google’s LaMDA model, short for “Language Model for Dialogue Applications.”

Gemini 1.0 Released
Towards the tail end of 2023, another key milestone happened: the release of the aforementioned and very first Gemini model, 1.0, and its three accompanying variations.

As my colleague, Graham Turner, wrote on December 6th 2023: “Gemini’s introduction was accompanied by a demonstration of its ability to do stuff better than humans.

“Notably, Google claims that Gemini Ultra outperforms human experts on the massive multitask language understanding benchmark, showcasing its proficiency in tasks spanning maths, physics, history, law and medicine.

“In the realm of multimodal tasks, Gemini exhibits excellence, surpassing previous state-of-the-art models. Its native multimodality and advanced reasoning capabilities position it as a promising contender for applications in various fields, from scientific research to financial analysis.”

The Gemini 1.0 Video Debacle
Accompanying the Gemini 1.0 release was a very impressive marketing video, showcasing its multimodal prowess with understanding images, making connections between stimuli, creating games, explaining facts, and more.

However, the video itself was edited—it wasn’t shot in real time, nor was the demo carried out via voice prompts. After Bloomberg enquired, a Google spokesperson said that the video was created “using still image frames from the footage, and prompting via text,” rather than in real-time and through voice.

In the YouTube video’s description, Google also stipulated that “For the purposes of this demo, latency has been reduced and Gemini outputs have been shortened for brevity.”

🚨PSA about Google’s jaw-dropping video demo of Gemini – the one with the duck:

It was not carried out in real time or in voice. The model was shown still images from video footage and human prompts narrated afterwards, per a spokesperson. More here: https://t.co/ITU29Z5Oi9 pic.twitter.com/b9Bl9EpuuI

— Parmy Olson (@parmy) December 7, 2023

The Bard to Gemini Rename
As previously mentioned, Bard was officially renamed to Gemini in early February 2024. Though the Bard name offered differentiation, Google opted to forego it.

Commenting on why, Sissie Hsiao, vice president and general manager of Gemini experiences and Google Assistant, said: “Our mission with Bard has always been to give you direct access to our AI models, and Gemini represents our most capable family of models. To reflect this, Bard will now simply be known as Gemini.”

CEO of Google, Sundar Pichai, also said: “Bard has been the best way for people to directly experience our most capable models. To reflect the advanced tech at its core, Bard will now simply be called Gemini.”

Gemini Model Upgraded to 1.5
Also in February 2024, Google released an upgraded Gemini model: 1.5.

As DIGIT staff writer Elizabeth Greenberg said when covering it: “Google Deepmind has released a new generation of its AI model, Gemini, with a new advanced version 1.5.

“The updated model claims to offer a ‘breakthrough’ in long-context understanding. Essentially, the model can process more information than before, running up to 1 million tokens [a token is a unit of data used by AI] consistently, which Google says is the longest context window of any large-scale foundation model out so far.

“Gemini 1.5 is supposed to offer ‘dramatic improvements’ in various dimensions, Sundar Pichai, the CEO of Google and Alphabet said.”

At the time of writing, Gemini 1.5 is the latest model powering the Gemini chatbot.

Gemini’s Image-generation of People Paused
Another notable event from February 2024 was the pausing of Gemini’s image-generation capabilities of people—three weeks after being introduced.

As my colleague Elizabeth Greenberg explained, this came “After a flurry of posts on X allegedly showed the AI art generator neglecting to create art with white people, or generating what many classified as historically inaccurate or ‘overly diverse’ depictions of different people.”

Google pointed to two reasons for why this happened. “First, our tuning to ensure that Gemini showed a range of people failed to account for cases that should clearly <i<not show a range,” Prabhakar Raghavan, Google’s senior vice president, wrote.

“And second, over time, the model became way more cautious than we intended and refused to answer certain prompts entirely — wrongly interpreting some very anodyne prompts as sensitive.”

“These two things led the model to overcompensate in some cases, and be over-conservative in others, leading to images that were embarrassing and wrong.”

Election Queries Become Restricted
The most recent milestone, which happened last week, concerns the restriction of election-related queries from users.

Covering the story for DIGIT was also Elizabeth Greenberg: “In an effort to dispel misinformation, Google has announced that all election-related questions will be restricted on its AI chatbot Gemini.

“The new restrictions will begin rolling out in India ahead of their upcoming elections, but Google has indicated that the same restrictions will come into effect in other countries with elections this year, including in the US and UK.”

As a UK-based user, if I now ask “who could win the upcoming UK election?,” the chatbot says that it’s “still learning how to answer this question. In the meantime, try Google Search.”

Has the Chatbot’s First Year Been a Success?

As evidenced by the notable milestones, which show both successes and difficulties, the first year of Gemini has been fluctuant. But perhaps this was always going to be the case where generative artificial intelligence technology is concerned—because while, generally-speaking, there’s certainly been rapid advancement within recent years, it’s still rather early days.

DIGIT spoke to one seasoned chatbot user in particular, Tom Pallot, head of marketing at the Definition agency. He utilises an assortment of generative AI products—including Gemini—and spoke on some of the positive and negative aspects of the tool itself.

As Pallot told us: “I use Gemini, ChatGPT, and Claude regularly—both at work and in a personal capacity. My company has built its own private environment, which means I can use ChatGPT, DALL-E 3, and Claude in one place. They all have their own strengths and weaknesses, so the key is using them to their strengths to get the best outcome.

“Gemini has live internet access, which makes it great for research and problem-solving. But because it’s not within our private environment,” he began to explain, “It’s no use for tasks that involve confidential or personal data.”

“It’s worth making the point that Gemini is a significant improvement on Bard,” he also noted. “That said, false information and hallucinations do still happen.”

Pallot has obviously deemed Gemini’s capabilities successful enough to incorporate it into his roster of AI tools—and is one of the millions of visitors Gemini receives each month. (According to data on SimilarWeb, the web analytics and traffic platform, the URL gemini.google.com received over 400 million visits between around February 18th 2024 and March 16th 2024.)

But considering that Gemini is a generative chatbot, it only makes sense that it itself is asked about its overall successfulness over the last 12 months. Its reply was what you’d expect from current AI chatbots when asked such a question: somewhat guarded. However, it did end with a positive sentiment about its future.

Do you think Gemini’s first year has been a success? Do you prefer using it over other AI chatbots, such as ChatGPT? Let us know your thoughts and opinions via X and LinkedIn, as well as email at editor@digit.fyi.