In a race against rival Microsoft, Google has presented a new version of its Internet search engine. It should answer questions with longer texts. The US group is also building the technology into other products.
In response to the use of artificial intelligence (AI) in the form of the text robot “ChatGPT” in the Microsoft search engine Bing, the market leader Google is following suit. Yesterday, Google introduced a new version of the search engine, which also uses so-called “generative AI” to deliver longer texts in response to open questions.
“We’re rethinking all of our core products, including search,” Google parent company Alphabet CEO Sundar Pichai said at the annual I/O developer conference in Mountain View, California. “Generative AI” is not only built into the new search engine “Search Generative Experience” (SGE), but also into the e-mail program Gmail, where it can deliver draft texts. At the same time, Pichai promises a prudent approach in order not to cause any damage.
Test run with Deutsche Bank
In the Google Photo application, the company wants the AI to change images, for example by placing people in the middle or coloring empty image areas. The search engine SGE should first be activated for a test phase in the USA for users, said the Vice President of the group, Cathy Edwards. There is a waiting list for this. Asked how the AI deals with misinformation online, Edwards said accuracy and citing trusted sources is a priority for Google.
The head of the Google Cloud division, Thomas Kurian, told the Reuters news agency that the new technology would also be offered to companies for creating reports, for marketing or for running chat assistants. Deutsche Bank, among others, could be won over for a test run.
Information can be wrong
Google has been working on applications based on artificial intelligence for years, but is currently under pressure. At the end of last year, the start-up OpenAI triggered a new competition in artificial intelligence – when it made its chat bot ChatGPT public. The software caused a stir because it can generate text. She is trained with huge amounts of data and estimates word for word how a sentence could go on.
Microsoft has integrated the app, which found new users at an unprecedented rate and has become for many the first standard version of “generative AI”, into its “Bing AI” search engine. However, this entails the risk that it may issue completely incorrect information.
Can “deepfakes” be prevented?
At the conference, Google assured that it would use the technology responsibly. “The only way to be bold in the long run is to act responsibly from the start,” said James Manyika, Google’s social responsibility for AI deployments. The group sees the danger that the software could strengthen prejudices or be used for the production and dissemination of false information. To protect against this, files created with the help of artificial intelligence, for example, should be provided with metadata so that they can be recognized immediately.
Google will also only provide verified developers with software that can automatically create dubbed versions of videos, Manyika said. With so-called “deepfakes” manipulated films are created with alleged actions of real people without the fakes being recognizable.
At the same time, Manyika emphasized that Google decided years ago not to make interfaces for facial recognition applications publicly available. When using artificial intelligence responsibly, everyone involved must work together.
Software formulates letters
Among other things, Pichai demonstrated how software can write letters. He also showed how users in Google’s photo app can not only remove unwanted objects and people, but also change their own position in the picture, for example. When you write a story, the software should be able to provide suggestions for further plot twists and automatically generated illustrations.
There are significant innovations in Google’s most important product – the Internet search. With the help of artificial intelligence, the search engine should be able to answer the question, for example, which of two nature parks is better suited for a family with children and a dog. The answers are formulated in complete sentences, as well as the usual Internet links.
Microsoft needs to save
The development and deployment of “generative AI” is very expensive. Microsoft has pumped billions of dollars into OpenAI and is heavily aligning its business around it. Employees also feel this on their pay slips: Microsoft announced yesterday that there would be no pay rises for full-time employees this year.
The company referred to the difficult economic environment. Along with the sweeping shift to AI, investments “in people, the business and the future” would need to be carefully considered.