We’ve come this far! Generative AI makes life more convenient, but also risks February 14th, 17:42
Taylor Swift held her first concert in five years in Tokyo. It is said to have attracted 220,000 people in four days from February 7th, exciting many fans.
However, in late January, a fake image of Swift spread on social networking sites such as the old Twitter account X, causing a major problem in the United States. It is believed that generative AI was used. Rapidly evolving generative AI is making business and our lives more convenient and significantly changing, but it is also increasing risks such as fake images, fraud, and leaks of personal information.
We followed the forefront of American generative AI.
(American Bureau reporter Daisuke Ezaki/Economy Department reporter Tomoka Miyoshi)
Fake image of pop queen goes viral
A fake image of Taylor Swift went viral online in late January. The images containing sexual content are believed to have been created using generative AI. Some of them were viewed 47 million times.
By January 29, SNS companies such as X were busy taking measures such as deleting fake images and restricting searches.
White House Press Secretary Jean-Pierre said, ``This is alarming. Social media companies are creating rules to prevent false information and sexual images of real people from being spread without their consent.'' It became such big news in the United States that people expressed concern that the virus was playing an important role.
Generative AI boom technology is rapidly evolving
Generative AI technology suddenly became well-known in November 2022 with ChatGPT, developed by the American venture company Open AI, and its use has expanded to the general public.
Many people were surprised by its ability to respond in natural sentences to characters and text input, but the evolution of technology continues to accelerate.
Multimodal AI, which simultaneously processes multiple data such as text, images, video, and audio, is currently a big trend in the United States.
In other words, within this trend, image generation technology was misused to create a fake image of Swift.
Generated AI extends to wearable devices
So, what kind of "multimodal AI" products and services are emerging in the United States?
One topic that has become a hot topic is the wearable device ``AI Pin'' that uses generative AI.
Orders are expected to begin in November 2023, with shipping scheduled to begin in March 2024.
The device is equipped with a camera and sensors, and is connected to the generated AI through wireless communication, allowing the device to accurately understand its surroundings and answer various questions.
A woman traveling in Taiwan picked up a dragon fruit and said to the device, "Can I eat this?" The camera recognized the fruit and the generated AI analyzed it.
She advises, ``Dragon fruit is good because it has low sugar content.''
Light was emitted from the terminal and icons were displayed optically on the palm of her hand.
For example, there is a convenient feature that allows you to control the music playback button with the palm of your hand.
The device was developed by startup company Humane.
The company was co-founded in 2018 by former Apple executives Imran Chowdhury and Bethany Bongiorno.
Mr. Chowdhury talks about the features of this product as follows.
Imran Chowdhury:
“Use the full power of AI wherever you are and seamlessly integrate it into your daily life. It embodies our vision.”
AI guides the generation of visually impaired people
The American startup Glidance has developed a device for visually impaired people.
At first glance, it looks like a cordless vacuum cleaner, but it has a camera and is connected to a generation AI.
If there is an obstacle in the direction of travel, the generated AI will judge the image captured by the camera and automatically guide the visually impaired person to avoid the obstacle.
If it cannot be avoided, it will stop suddenly and the handle will vibrate to let you know.
In fact, I experienced it myself, and it gave me the impression that I was guided quite accurately and could walk easily.
Generation AI in cars too
Rapidly evolving "multimodal AI" is beginning to be incorporated into the automotive field.
``Sony Honda Mobility'' is a collaboration between Sony, a major Japanese electronics manufacturer, and Honda, a major automobile company, to develop EVs. Together with American IT giant Microsoft, they are developing a system to be installed in cars that utilizes generative AI. We announced in January 2024 that we will move forward with this.
The new EV is characterized by proposing new ways to enjoy the inside of the car, such as videos, music, and games.Through dialogue with the generated AI, it is possible to not only operate the car but also receive suggestion-based communication from the AI. I am assuming that.
Some media outlets are already fantasizing about the details of new services, such as AI that listens to conversations and automatically performs services or becomes a chat partner in the car.
Jessica Hawke, Microsoft
: “As the AI landscape rapidly changes and new technologies and innovations emerge in the mobility industry, safe and responsible AI remains a top priority.”
Izumi Kawanishi, President of Sony Honda Mobility
: ``AI technology is evolving rapidly. Until now, cars have had a one-way relationship of being operated and driven by people, but with the evolution of mobility, cars are becoming more human-friendly. I think we can create a two-way relationship.I want to rethink the relationship between people and cars."
Detect “deep fakes” using AI that generates them
On the other hand, as technology evolves, risks are also increasing. In the United States, there is already a growing movement to turn generative AI risk countermeasures into a business.
Gartner, an American research company, analyzes that risk management of generated AI will become important in 2024.
▽AI Trust
▽Risk and Security
▽Management
The keyword is "AI TRiSM," which is an abbreviation of the initial letters.
There is a company in New York that is already developing this "AI TRiSM" as a business.
This is a startup company called "Reality Defender."
It is known as a company that develops technology to detect whether fake images and sounds generated by AI are fake.
Too similar!
CEO Ben Coleman conducted an experiment in which he created a fake version of my voice and applied it to his company's system.
First, I created a fake voice based on the audio data I recorded on my smartphone.
You can have a fake voice read the text, but I was surprised because it sounded so similar to my own voice.
When I ran my fake voice through the system, the red part, which indicates that it was a fake voice, accounted for most of the black part, which shows the volume of the whole voice, and 98% of the voice was created. It was determined that.
By using generation AI to detect the extremely minute changes that occur when converting a text into audio, it is possible to determine whether the audio is fake or not.
It seems that these minute changes cannot be detected by the human ear.
It is also said to be able to identify fake images.
This video, which at first glance appears to be a statement from President Putin, can be easily identified as fake using this company's system.
It detects slight discrepancies between the mouth movements and the voice when speaking, and determines that the image is a fake.
Ben Coleman, CEO
: "Fake audio, images, and video are evolving at an extremely rapid rate. Every day, our adversaries are creating new tools to create fake faces and voices. It's what keeps us up at night. We must constantly update our technology and offer new models to help our customers identify fake faces and voices in real time.”
Generation AI risk monitoring business
Furthermore, businesses that provide companies with countermeasures against the various risks they face when using generated AI in business are also attracting attention.
I visited Robust Intelligence, a startup company founded by AI researchers from Harvard University, in San Francisco, which is home to many technology companies.
For companies, we provide a service that uses generated AI to monitor the risk of providing incorrect information to customers or leaking personal information.
For example, in the case of a company that introduces a system that uses generative AI to answer questions from customers via chat etc., if the system tries to convey incorrect information to the customer, the AI that has read the company's vast amount of data etc. The software detects this and immediately informs the company to prevent it from happening.
The risk of generative AI is not just the spread of false information.
For companies with large amounts of personal data, leakage of personal information is a major risk.
Furthermore, for companies that are expanding their business globally, it is also important to be able to comply with the regulations on AI set by each country's government.
The company believes that demand will increase in the future for businesses in the so-called "defense" field, which require specialized skills and knowledge.
Yukito Oshiba, Co-Founder, Robust Intelligence
: “AI risks are increasing, and new vulnerabilities emerge with each new technology. That’s why we need specialized organizations and companies focused on problem solving.”
Two things that felt revolutionary in my life
In March 2023, Microsoft co-founder Bill Gates posted a blog titled ``The era of AI has begun.''
With the subtitle, ``Generative AI is as revolutionary as mobile phones and the Internet,'' he wrote, ``In my life, there have only been two demonstrations of technology that I felt were revolutionary.'' He points out that the GUI was the forerunner of current operating systems including Windows, and the other is the generation AI of Open AI.
It's no wonder that it's evolving so incredibly rapidly, as it's one of only two technologies Gates thought was revolutionary.
It is still unclear what kind of convenience this accelerated evolution will bring us, and what kind of economy and society it will shape.
At the same time, the risks may exceed our imagination, and I strongly feel that society as a whole needs to think about how to use them, including the establishment of rules.
(Broadcast on "BS Kokusai Hoso" on January 12th)
Daisuke Ezaki
, reporter for America General Bureau Joined
the agency in 2003 After working in
the Miyazaki Bureau, Economic Affairs Department, and Takamatsu Bureau, he is currently at his current position.
Economic Affairs Department reporter
Tomoka Miyoshi Joined the Bureau
in 2017
After working at the Nagoya Bureau,
she is currently in charge of the IT and electrical industries.