Google launched Bard early last year as an answer to the generative artificial intelligence craze, packed with the ability to hold a conversation, generate code, and solve logical problems. Bard's biggest advantage over competitors over the likes of ChatGPT was that it could access the internet, though it managed to falter behind the rest of the competition in several pretty crazy ways. Since then, it's undergone a ton of improvements and followed in Microsoft's footsteps by rebranding. Google Bard is now Gemini, and it's powered by the Gemini language model underneath.
Gemini uses a whole new language model
PaLM 2 is gone
Google Bard was in a bit of an odd spot when it came to the technology behind it, as Google hadn't been particularly clear about the language model underpinning it. Google announced at last year's I/O that Google Bard would be using the PaLM 2 language model to generate text, but asking Bard what it uses will yield an answer stating that it makes use of LaMDA for dialogue and PaLM to generate text. No mention of PaLM 2 whatsoever. Having said that, AI isn't exactly the most trustworthy when it comes to factual claims. Google said that Bard uses PaLM 2, so we operated under the assumption that it used PaLM 2. Palm stands for Pathways Language Model, and this was the second iteration of it.
Since then, though, Google has entirely overhauled Bard. It's now called Gemini, it's now integrated into the Google Assistant on Android, and there's a paid premium tier called "Gemini Advanced" that'll cost you $20 a month. This is akin to what the likes of Microsoft has done with Copilot Pro, and initial results suggest that Gemini is a significant improvement over PaLM 2. What's more, Gemini Advanced comes with the Gemini Ultra, and you'll get a bunch of other perks that come with Google's "AI Premium" tier of Google One, such as 2TB of cloud storage.
Gemini still retains a major advantage that it had over the free tier of ChatGPT too in that it can access the internet, meaning that you can feed it a URL and ask it to summarize the page for you or even ask it to look something up for you. Copilot can do this as well, though, so Google isn't exactly in a league of its own in this regard anymore.
Gemini was made by Google
It's a collaborative effort
Google's Gemini was developed by Google and released on March 21, 2023 as Bard. It wasn't available for everyone at launch, expanding to more than 180 countries as of Google I/O, though excluding EU nations. However, Google then struck a deal with European regulators, allowing for its rollout on July 13, 2023. It's now been rebranded to Gemini, and Gemini is available globally.
Gemini is a collaborative effort between Google AI and DeepMind, and the project is overseen by Jack Krawczyk. The Responsible AI team is also heavily involved in the process, and their job is to ensure that AI is developed thoughtfully and safely.
What can Gemini do?
All of the same, but better
Gemini can do everything the free version of ChatGPT can, with some extras on top. One of its biggest advantages is its live access to the internet, as you can feed it information from a live webpage and gain an insight into its contents. That's something that ChatGPT can't do right now. As well, replacing Google Assistant means it'll soon be integrated into every Android phone out there, giving Google a massive edge. Aside from that, you can program, research, summarize, and write with the help of Gemini.
However, LLMs aren't perfect. Google says that "if you ask an LLM to solve a mathematical word problem, it will predict an answer based on others it’s learned from, not based on advanced reasoning or computations. To this end, we have seen Bard present responses that contain or even invent inaccurate information." That's still true even now from our early testing, but that will likely improve. In contrast, ChatGPT Plus will use a Python interpreter to run code and calculate the answer for you.
Some of the best features of Gemini include the ability for it to respond with images, view other drafts generated, export text into multiple formats, and code explanations. Most of this the free version of ChatGPT can't do, certainly without plugins.
What is Gemini Advanced?
The premium tier of Gemini will set you back $20 a month
Gemini Advanced, which utilizes the Ultra 1.0 AI model, is an enhanced version of Gemini that Google claims outperforms the standard model, especially in dealing with complex tasks. Google highlights several areas where Gemini Advanced excels, including coding, logical reasoning, interpreting detailed instructions, assisting in creative endeavors, and grasping contextual nuances.
Google claims that the Ultra 1.0 AI model surpasses human capabilities in extensive multitask language comprehension, covering 57 subjects such as mathematics, law, and ethics. Additionally, users of Gemini Advanced will gain early access to forthcoming and exclusive AI tools, including Google Photos features and 2TB of cloud storage. This offering goes beyond a mere AI model, unlike ChatGPT Plus, integrating seamlessly with Google's suite of applications like Drive, Sheets, and Slides. This positions it as more of a direct rival to Copilot Pro.
Google Gemini Advanced vs ChatGPT Plus: Which is better?
Both services are great and cost the same, but which is better: Gemini Advanced or ChatGPT Plus?
How to use Google Gemini
All you need is an account
You can use Google Bard by navigating to Google's own landing page for it. You'll need a Google account, but that's more or less the only barrier to entry. If you live in an unsupported country, you will also be able to use a VPN to gain access to the service, which is how I used to use it despite living in Ireland.
Once you're in, you can ask it basically anything. Here are some things I've asked Gemini to do so far:
- Make me a meal plan
- Make me a fitness plan
- Help me with coding
- Write a poem
- Research for me
It can do anything you can think of to ask it, so play around and ask it different things!
The future of Google Gemini
Massive leaps for Google
As we say goodbye to Bard, it's clear that Google's commitment to competing with Microsoft and OpenAI is starting to pay off. The company has posed itself as massively concerned with the responsibility aspect of AI, but to many, that came across as an excuse for falling behind the rest of the competition. With a different language model and massive improvements already being made, competition is obviously a good thing.