I knew Google was planning to make big AI announcements at I/O 2024. But then, rumors spread last week that OpenAI would unveil a Google Search rival for ChatGPT. Sam Altman dispelled those reports when OpenAI announced a ChatGPT event for Monday, just one day ahead of Google’s big Gemini fest. OpenAI said it would discuss ChatGPT advancements and offer AI product demos.
One thing was immediately clear: We were about to witness another significant battle in the ongoing AI war.
Fast-forward to today and we now know everything there is to know about the AI updates that OpenAI and Google brought to the table. ChatGPT got its big GPT-4o upgrade, while Google brought variants of Gemini AI to virtually all its existing products. It also unveiled a new Project Astra AI assistant for the Gemini app that matches GPT-4o’s multimodality.
So who won? Well, it’s complicated.
OpenAI and Google are spearheading this ongoing AI revolution. Each has strengths and weaknesses, even after Monday and Tuesday’s announcements. But I do think the real winner is the user. Whether you choose GPT-4o or Gemini, you won’t be sorry. Also, I think we’re heading to an era where you might prefer one AI system over another, but many people will end up using multiple AIs from different companies.
Why GPT-4o won
Before I even saw Google’s new Gemini advancements, I had to appreciate how amazing the GPT-4o upgrade for ChatGPT really is. OpenAI pulled off something Google wished was possible back in December when it had to fake its first Gemini demos.
GPT-4o gives ChatGPT eyes and a voice, and the chatbot can use these “senses” while talking to you. These abilities will change the way you look at the world.
OpenAI’s GPT-4o handles voice conversations almost like a human. You can interrupt the AI when you want to adapt the conversation without losing the context. GPT-4o also supports images and videos, and it answers questions about what it can see in real-time.
Also important is the fact that GPT-4o is faster than GPT-4 when answering prompts. It’s not all about voice input, as text prompts will generally get you faster results. The new model should offer better reasoning than previous ChatGPT models.
GPT-4o is available right now to ChatGPT users on the Free and Plus tiers, with the latter getting higher message limits. The upgrade is also available on mobile, with a Mac desktop app launching soon.
ChatGPT also received a major privacy upgrade in the days leading up to the Monday event. Now, you can prevent your data from training the AI without losing your chat history. That’s a big deal, especially considering what the GPT-4o upgrade will make possible.
Why Google Gemini won
I might personally use ChatGPT a lot more than Gemini, but there’s no question that Google made a big AI statement of its own at I/O. I dare say I/O 2024 is even bigger for Google than last year’s show, which was also focused on AI developments. It felt like a tour de force from Google this year, as the company demoed the various ways it has improved Gemini.
Gemini 1.5 Pro is Google’s best model, which can handle prompts of up to 1 million tokens. That limit will double to 2 million soon. It’s already available to users who pay for the Advanced version of Gemini. This is one area where Google seems to outperform OpenAI, though Gemini users on the free version won’t be able to take advantage of the 1 million-token limit.
Google also showed a prerecorded demo of Project Astra. Like GPT-4o, this is a multimodal AI agent that can talk to you while “looking” at the world around you through a phone’s camera. The conversation seemed fast and fluid, with the AI adapting quickly to answers. But again, this wasn’t a live demo, so we don’t know how Astra will perform in real-life situations.
Still, one of Project Astra’s mind-blowing features was its ability to retain memories of what it had just seen. For example, it remembered where the user’s glasses were when she asked about them. But again, unlike GPT-4o, this wasn’t a live demo. Google’s tech isn’t ready to be used widely right now. It’ll be a few months before it reaches the Gemini app.
The real strength of Gemini is that it’s part of a much larger ecosystem than OpenAI can offer for ChatGPT. Google adapted Gemini to work with various products, such as Gmail, Google Workspace productivity apps, and Google Search.
More importantly, it’s building Android around AI. That includes upgraded Circle to Search functionality that can even do your homework and a Gemini app that’s aware of context. The app can see other apps on your phone, just like the GPT-4o will see apps on your Mac.
Which AI will I use the most?
There’s no question for me: ChatGPT will remain my primary genAI product for the foreseeable future, and I’ll rely on GPT-4o for my prompts. I’ll use it in the wild more and take advantage of voice, images, and video features.
At the same time, I’ll probably use Gemini AI inside certain Google apps. I might also try Google Search and explore Gemini on phones, especially once the new features roll out. Project Astra is also on my radar.
Again, I’m saying that people likely won’t have to stick to just one generative AI product.
ChatGPT with GPT-4o and Google Gemini each promise great experiences. Some are available right now, and some will come soon. OpenAI and Google will likely try to copy each other’s AI initiatives in the coming months and years to compensate for weaknesses. But you don’t have to worry about being locked into one platform or the other, like so many people are with mobile and desktop operating systems. AI will be available across platforms, and you can use different options to meet different needs.
