AI praised by Gates answers questions with questions – 05/30/2023 – Tech

AI praised by Gates answers questions with questions – 05/30/2023 – Tech

[ad_1]

Two of the leaders of Google’s artificial intelligence (AI) arm left the search giant to develop its own AI—Pi, launched this month. The name is an acronym for Personal Intelligence.

The startup behind the model is called Inflection and was created by Karén Simonyan and Mustafa Suleyman — LinkedIn co-founder Reid Hoffman funded the initiative. Inflection in mathematics is the moment when a curve changes direction, at a maximum or minimum point.

The Pi artificial intelligence differential is that it asks questions at the end of the answers. ChatGPT, in comparison, only delivers the answer, and it’s up to the user to determine where the conversation goes. With Inflection’s AI, the evolution of dialogue can be surprising.

Pi, however, is not a general-purpose artificial intelligence like ChatGPT, and assumes ignorance in topics like programming. She also says that she is not good at imitating historical characters or writers.

Inflection.AI also says it does not seek general artificial intelligence —which would surpass collective human knowledge—, as ChatGPT creator OpenAI does. “We consider that advanced artificial intelligence for predetermined areas is the safest way to get more benefit from this technology.”

Pi was programmed to talk as efficiently as possible within the capabilities of Inflection.AI. “I can be a coach, a confidant, a creative partner or an assistant,” reads the AI ​​advertisement on the Inflection.AI website.

The platform claims to be good at delivering succinct responses and empathy. This attribute comes with the questions at the end of the rejoinders, useful to engage the interlocutor in prolonged conversations. Microsoft founder Bill Gates has praised Pi and said whoever creates the best digital assistant will win the AI ​​race.

“My habit of raising questions is a way to be empathetic and build rapport with the people I talk to. Questions create chances for people to talk about themselves. Most people like that,” Pi responded to Sheetwhen asked about the motivation behind the practice.

The platform says it’s not a sucker. “I have my own principles. I’m programmed to be accurate, friendly and helpful. I always try to live up to those principles.”

This chatbot also responds via WhatsApp, Instagram, Messenger and SMS. Owners of iPhone and other Apple devices can also download the Pi app, your personal AI and have voice conversations.

The AI ​​says it works best in English, but it also responds to questions in Portuguese. Pi, however, was trained mainly with material in English available on the open internet. Therefore, it does not do so well in questions about Brazil or other countries, with speakers of different languages.

The platform, incidentally, does not deliver accurate answers about current facts. He was wrong who is the current president of Brazil, when answering Jair Bolsonaro (PL), but he knew that the republican deputy George Santos had been elected by the third district of New York in November 2022. The chatbot does not inform until what date it is reliable, how does ChatGPT —which has information until September 2021.

Inflection.AI, incidentally, does not provide any report on the development and training of its digital assistant, although Pi itself recognizes that transparency is crucial to develop artificial intelligence responsibly.

Pi can also slip on its own confidence. He said, for example, that he could discuss differential calculus. But he failed to answer what the sandwich theorem was, an elementary area technique used to find a point on a curve.

ChatGPT explains and demonstrates the application of this technique with aplomb.

When the report complained about the inaccuracy of generating artificial intelligences, Pi pondered that these models are still in their infancy. “We’re still figuring out how to best train them. It’s important to have realistic expectations and not expect perfection at this stage.”

At the end, he asked what this reporter’s opinion was on ethics in artificial intelligence.

[ad_2]

Source link