Why Google decided to hit the brakes with ingenuity

Google announced on Wednesday (11) a series of novelties focusing on the technology combined with its services, including the new Android. In terms of intelligence, however, the theory remains that the American giant had set foot on the attack. Reports linked to this area of ​​technology focus on activities to test and correct known problems.

For example, Google has given other developers access to its Monk Scale, a list of 10 skin tones created by Professor. Ellis Monk, from Harvard University. The idea is that the measure will help train AI models so that technology recognizes (and respects) human differences, reducing the risk of discrimination – an issue that has in this market for many years.

Google has also announced a new app called AI Test Kitchen, which will allow people to try out the latest technology models developed by the company and find lines and provide feedback. them before they are presented to the public.

Zoob GhahramaniThe vice president of research for the AI ​​division, believes that the acquisition of intellectual property will be slow and slow because there are still many issues to be solved, and Google needs to be careful more.

Fixed past (and current) errors

The new move will address counter-insults from the education community that Google has already targeted. During the development of new terms for intellectual property, even the company’s own employees have found dissatisfaction.

There have been complaints from employees who say they have been fired for reporting issues, such as racism, gender discrimination and racism, in the guidelines. of the company.

An example is former head of research in AI ethics, computer scientist Timnit Gebru. His dismissal comes after he accused the company of racism and censorship. At the time, Google claimed that there were “too many speculations and misunderstandings” about the release.

The AI ​​Test Kitchen app will be available for Android, but it will be on invitation to be installed on the cell phone. The application will test LaMDA 2, a unique AI model in human language that Google is developing.

It works simple: you speak, your way, and it responds, trying to understand the nuances and subtleties of words that people are used to, but that can be difficult for robots to interpret.

According to Google, the app will be a test site for the company to test products developed with LaMDA 2, such as Research, for example, community feedback to improve things. shipments – and, of course, address the prejudices that may arise.

Invitation to download AI Test Kitchen will be restricted. Probably because when big companies publish intelligent technology, without proper proof, the benefits can be devastating, as has already been seen.

In 2015, a black couple registered as “gorillas” on Google’s image service. And you may remember the problem Microsoft had with Tay, its AI-powered chat bot, after Twitter users “reported” it to create racism and slander.

Or even ask Delphi, an expert designed to solve ethical problems that can persuade people to avoid murder.

Google’s new app has invited testers in the community to criticize its products, but more so, control the comments. This shows that the company expects that something is still going wrong and needs to be tweaked.

Future use of AI

The AI ​​Test Kitchen app offers three types of experience:

  • “Imagine”,
  • “About”
  • “Name”.

Each of them tried different functions of the language developed by the company:

Consider: You can sign up for the real place or think about it, then LaMDA will try to explain it. The system should be able to explain everything in detail.

  • For example, when typing “Think I’m under the sea”, the AI ​​responds with “You are in the Mariana Trench, the deepest part of the ocean. The waves crash on the walls of your submarine . surrounded by all darkness. “

Talk about: Artificial intelligence will try to hold conversations on all topics. The idea is to see if the system is no different from what is being discussed. In the major tech examples:

  • The AI ​​asked, “Have you ever wondered why dogs like to play with lifts?”
  • The user simply answers “why?”.
  • The system understands the content and explains that there is something wrong with the dog’s odor.
  • If the user asks “why do they smell better?”, The system understands (or should understand) that “they” are about dogs, not people should repeat key points.

Name: The app tries to list all the subjects in the relevant topics.

  • When you say, “I want to grow vegetables”, the answer would be “What do you want to grow?” and a list of operations and collection items, such as “water and other maintenance.”

Leave a Comment