This week, both Google and OpenAI announced that they have built "super" artificial intelligence assistants. These tools can converse with you in real-time, allowing you to interrupt them at any time, and they can also analyze your surroundings through live video and provide instant translation of conversations.

On May 13th local time, OpenAI showcased its latest flagship model, GPT-4o, for the first time.

In the live demonstration, it was able to read bedtime stories and help solve math problems, with a voice that sounded like the artificial intelligence girlfriend of Joaquin Phoenix in the movie "Her."

Apparently, Sam Altman, the CEO of OpenAI, has been keeping this point in mind from the movie.

On May 14th local time, Google announced a series of new artificial intelligence tools, including a conversational assistant called Gemini Live, which can do similar things as GPT-4o.Google also revealed that it is building an "omnipotent" artificial intelligence agent, which is currently under development, but will not be released until later in 2024.

Soon, you will be able to explore these tools on your own to see if they will be used in daily life as developers hope, or if they are more like those little tricks that will eventually lose their appeal.

Advertisement

Here is what you should know about how to access these new tools, the purpose of using them, and the associated costs.

OpenAI's GPT-4oIts function: This model can converse with you in real time, with a response delay of about 320 milliseconds, which OpenAI states is on par with natural human conversation.

You can ask the model to explain anything you capture with your phone's camera, and it can assist you with tasks such as writing code or translating text. It can also summarize information, generate images, fonts, and 3D renderings.

How to access: OpenAI has indicated that it will begin rolling out the text and visual capabilities of GPT-4o in web pages and GPT applications, but the date has not yet been announced. The company has stated that it will add voice capabilities in the coming weeks, but a specific date has not been determined.

Developers can now use the text and visual features through the official API, but the voice mode will initially be available only to "a small number" of developers.

Cost: The use of GPT-4o will be free of charge, but OpenAI will set usage limits, and users can increase these limits through a subscription.For those who join the OpenAI subscription plan (starting at $20 per month), the message capacity of GPT-4o will be increased by five times.

Google's Gemini Live

What is Gemini Live? This is a product from Google that directly competes with GPT-4o, which you can have real-time conversations with. Google has stated that later in 2024, you will also be able to communicate with this tool through video.

The company promises that it will be a helpful conversational assistant for preparing for interviews or practicing speeches.How to Access: Gemini Live will be joining Google's premium artificial intelligence program, Gemini Advanced, in the coming months.

Cost: Gemini Advanced offers a two-month free trial period, after which the monthly fee is $20.

So, what is the Astra project? Astra is a project to build an omnipotent artificial intelligence agent. Google demonstrated the project at the I/O conference, but it will not be released until later in 2024.

Google DeepMind's Vice President of Research, Oriol Vinyals, told MIT Technology Review that people will be able to use Astra through smartphones and desktop computers, but the company is also exploring other options, such as embedding it in smart glasses or other devices.

Which one is better?As of now, we are unable to experience the full versions of these models, making it difficult to determine which is superior. Google showcased the Astra project through a meticulously crafted video, while OpenAI opted for a seemingly more authentic live demonstration with GPT-4o.

However, in both cases, the models were asked to perform tasks that developers may have practiced many times. The true test will come when they are first exposed to millions of users with unique needs.

That being said, if you compare the video released by OpenAI with Google's video, these two leading tools appear strikingly similar, at least in terms of ease of use.

Overall, GPT-4o seems to have a slight edge in the audio domain, demonstrating realistic sounds, conversations, and even singing. Astra, on the other hand, showcases more advanced visual capabilities, such as being able to "remember" where you left your glasses.

OpenAI may roll out new features more quickly, which means its product could initially see more usage than Google's, as Google is not expected to fully launch its product until later in 2024.It is too early to determine which model generates "hallucinations" or false information less frequently, and which model can produce more useful responses.

Are they safe?

Both OpenAI and Google have stated that their models have been well-tested. OpenAI says that GPT-4 has been evaluated by more than 70 experts in the fields of misinformation and social psychology.

Google says that Gemini "has the most comprehensive safety assessment of any Google AI model to date, including bias and toxicity."But these companies are building a future where artificial intelligence models search, review, and evaluate real-world information to provide us with answers to our questions. Compared to relatively simple chatbots, it is wiser to be skeptical of the information they tell you.