Amid a wave of new hardware, including Pixel 7, Pixel Buds Pro and the new Pixel Tablet, Google turned down a development at its I / O developer conference that went largely unnoticed: its AI can now understand jokes.

Jokes, sarcasm and humor require an understanding of the intricacies of language and human behavior. In order for a comedian to say something sarcastic or contradictory, the audience can usually distinguish the tone and understand that it is rather an exaggeration, something that has been learned from years of human interaction.

But PaLM, or the Pathways Language Model, taught him without being explicitly trained in humor and the logic of jokes. After receiving two jokes, he was able to interpret them and spit out an explanation. IN blog postGoogle shows how PaLM understands a new joke that has not been found on the Internet.


Example of Google’s PaLM learning model for understanding and explaining a new joke.


Understanding Dad’s jokes is not the ultimate goal of Alphabet, Google’s parent company. The ability to analyze natural language nuances and queries means that Google can get answers to complex questions faster and more accurately in more languages ​​and nations. This, in turn, can break down barriers and keep people from communicating with machines through predefined means, and instead interact more seamlessly. This may involve answering questions in one language by finding information in another or writing code to a program while one speaks in the model with a specific task.

The power of this technology seemed lost to some on Twitter. When searching for # GoogleIO2022, the best results focus on Pixel hardware. Indeed, to make people understand the power of their artificial intelligence technologies, Google plans to do so with custom chips housed in custom Pixel devices.

“How do you tell my mother, LaMDA or PaLM, not to mention natural language processing,” said Tuong Nguyen, senior chief analyst at Gartner, a technology research and consulting firm. “The devices serve in part as a delivery mechanism for all the amazing things they do.”

Google can combine this with practical examples of how PaLM can eliminate small troubles. Speaking on stage at Google I / O on Wednesday, Alphabet CEO Sundar Pichai cited the frustration faced by people who speak lesser-known languages ​​when trying to find answers to their language problems. . The answer is probably online, but it will probably be in English or Spanish.

Pichai set an example when PaLM was asked in Bengali what the popular pizza toppings are in New York. The model can find the answer in English and translate it back to the user in Bengali.

“One day we hope to be able to answer questions on more topics in every language you speak, making knowledge even more accessible on Search and across Google,” Pichai said.

Do you surpass Star Trek data?

PaLM is Google’s largest AI model to date and trained on 540 billion parameters. Can generate code from text, answer a mathematical word and explain a joke. He does this through a prompt from a chain of thought that can describe multi-stage problems as a series of intermediate steps.

On stage, Pichai described him as a step-by-step teacher to help the student understand how to solve a problem.

If what Pichai said on stage is correct, Google has essentially surpassed Star Trek and 400 years of fictional AI development, as seen by the character Data, who he never understood the intricacies of humor. What’s more, it looks like Google has caught up with TARS from the 2090 film Interstellar, AI, which is so good at humor that Matthew McConaughey’s character told him to tune it.

PaLM’s ability to understand humor and draw logical conclusions helps Google solve new challenges that would have previously taken someone with specific experience.

In that sense, Google’s AI is a time machine. AI models can complete years of training and research in seconds. Time-machine companies that can connect people faster or instantly are the ones that dominate global markets and change lives. The value of these companies for human interoperability is enormous.

Google, which is primarily an ad search and sales company, can use its technology to ensure that people continue to turn to Google to answer their questions. This, of course, will give Google more data to target advertising to customers. Companies can also turn to their servers and AI to solve complex problems. At I / O Pichai announced a publicly available data center and machine learning center worth $ 9.5 billion in Mayes County, Oklahoma. There, Google Cloud customers can use the facility’s nine computing power exaflops to launch sophisticated models and help solve problems in medicine, logistics, sustainability and more.

As for the consumer, Google is investing in custom development with its Tensor chip line for its Pixel phones.

“Google is essentially a data, AI and search company. And they provide revenue from your attention and data, “said Avi Greengart, president and lead analyst at Techsponential. “Hardware is an extension of the platform and the ecosystem that drives it all.”

In order for Google to fulfill its vision of “computational calculations” or a future in which people can use computers so intuitively that they don’t realize they are using it, it had to invest in its own hardware.

“Another company may not be talking about something coming in the fall, when we’re not even in the spring because you’re afraid of getting burned,” Greengart said. “Google is not afraid of that. They would rather build on that vision of having interoperable software that does great things. ”

Previous articleMusk suspends Twitter acquisition “temporarily”
Next articleHow corporate risk management is changing