Google’s Bard Is Now Gemini - What’s New?

Alright, it is here. Bard is now Gemini. Google Smart Assistant AI has been updated big time. When they unveiled Gemini they showcased three Nano for efficiency Pro that was added to Bard that we could use for free and Ultra. The Ultra was the super interesting part. Their paper discussed that it is so good they think it might even give GPT4 a little competition. And here comes the best part. Oh yes, we can try it for free right now. So where can we use it and what can it do? Let's find out together. Dear fellow scholars, this is two minute papers with Dr. Zsolnai. If you have an Android phone you can opt in through Google Assistant or download the Gemini app and there you go. And if you have an iPhone, it will come to the Google app where you just hit a toggle and and it will be right there. As of the making of this video it was not available for me but they announced that it is coming in the next few weeks. Or you can go to gemini.google.com and just use it from the web. The regular Gemini is pretty smart and remains free. And Gemini Ultra is called Gemini Advanced here and is available as a paid subscription. And there is a free trial where all of you fellow scholars can try it for two months and you also get two terabytes of storage. Google has some advantages here as it can integrate with your Gmail account, can summarize your messages and write emails on your behalf. It also has access to your Google Docs slides sheets and get information and do work on them. It can also help you with your coding, can evaluate your ideas and more. It can do almost anything that you would expect from a smart assistant that works for you and it is going to be right in your pocket all day almost all around the world. Initial experiments show that it can also estimate the amount of calories from your lunchbox label and so much more. I also like that it can try to verify the validity of its statements from other sources. This is a feature that has existed before, but now I sometimes see it do it on automatically. Really nice. Love it. And like OpenAI's Dall E3 this can also draw some beautiful images for you too. More competition in that area as well. It will be hooked into your Google Assistant as well, making it so much smarter. This is also great because it will understand the context from your previous messages. Something that all major assistants struggle with these days. As far as I understand, Gemini is deployed to 230 countries and understands more than 40 languages. According to their press release, Gemini Advanced is available in 150 countries and currently only speaks English well. I tried German, Japanese and more and they seem to have worked. And here comes the really cool part from the earlier paper. We saw plenty of benchmarks that showed how amazingly good the Ultra is for more difficult tasks like coding, logic, reasoning. But we could never try it ourselves and now we can. We can now rerun all of these experiments and even run our own. This is heaven. Also, competition against OpenAI is heating up that is also excellent for us, the scholarly users. I myself ran an experiment from the USA Computing Olympiad, a task that they call insanely difficult. Yes, this is the one where you have to assemble the robotic cows, one of my favorites. This task requires a technique called dynamic programming and the size of the problem is huge, so the implementation has to be really efficient to have even a fighting chance to solve it. This is something that GPT4 was able to solve correctly. So let's see. Gemini Ultra. Hmm. It starts in the right direction, but unfortunately not yet. Now when GPT4 arrived there was a huge 155 page long paper that subjected it to every single test that you can imagine and some that were really solving mazes, mathematics, playing text based games, drawing maps, writing a full computer game and more. And I am sure that the same will happen to Gemini Advanced as well and I cannot wait to see them. I tried to look around and see what you fellow scholars are up to and my first impression seems to be is that GPT4 likely remains the smartest AI assistant for now, while Gemini will have the advantage of being able to work within Google's ecosystem and connect to your other apps a little better. So what experiments did you run? Did it perform well? Did it not? Is it as good as it seems? I'd love to know what you fellow scholars are up to. Let's run some more experiments and let me know in the comments below how they went until the new papers appear and they are coming soon so we'll know so much more. I am really excited for this. If you are too, make sure to subscribe and hit the bell icon to not miss out on them. What a time to be alive. Experiment tracking, Model evaluation and production monitoring for your deep learning projects and LLM apps. This is what weights and biases does and it is the best everyone is using it. Try it out now at WNB Me Papers or click the link in the description below.