It looked like the incubator of the Silicon Valley, with specific glasses and headsets of Ted-tot-style.
On Monday, February 10th, Google CEO took the stage at the artificial Wisdom Action Conference in Paris. From that platform of the Grand Palace, he began a new golden age of innovation.
“We have added more than 5 new languages allegedly allegedly by 5 million people worldwide using the AI strategy,” this technical senior said, “We have added the total number of 249 languages, including 60 African languages- and added,” this technical senior said.
However, many of his monotonous presentations did not receive many reactions, including many world leaders, researchers, non -governmental organizations and technology executives.
However, for artificial intelligence, the words of the beautiful Pichai were a symbol of a quiet victory. The victory that was achieved in the mysterious world of digital diplomacy after two years of intensive conversation behind the screen.
In New York, Joseph Nakulbhaul Nagula, a digital policy adviser to the international organization on the UN mission, says, “It is clear that this message has reached and listens to technical agencies.”
Linguistic
However, in the beautiful Pichai speech, the primary generator was not mentioned in AI’s linguistic mistakes. Early generator AI is a branch of artificial intelligence, which is capable of creating all kinds of key materials, from writing pictures, music and animations.
When the OpenAI launched the chat GPT in 2022, non-English-speaking people soon realized its boundaries.
For example, there is a detailed, informative answer for any question asked in English. If you ask the same question in French you will only get the answer to two paragraphs. And then you will find an apology: “Sorry, I didn’t train about it,” or “my model has not been updated after this date.”
This gap is because the complex equipment of AI equipment, to digest huge internet data, depends on large language models like GPT -4, meta Lalama or Google’s Gemini, which helps them give them proper answers as questions.
However, the Internet itself is mostly English language. And although only 20 percent of the world’s population speaks in English at home, about half of the big AI model training data has been done in English.
Currently, Chat GGPT has seen the answers to French, Portuguese or Spanish, but they are still far behind in their English parts.
Need to pay further attention
Joseph Naklavo Nagula said, “Most refreshing information is available in English. So, AI models are mostly published for thoughtful, trained and English language. As a result, other languages have to fight to reach this level.
This section is not just numerical. In the absence of training in any language, AI reaches the wrong or confused answer to an “AI hallucination” – just as your very confident friend responds to a social event.
In response to the request of a famous person’s biography as a reasonable example of AI hallucinations, he can discover or introduce the Nobel Prize in a strange parallel career. Some similar answers were published from Chatzept at this request by UN News:
UN News: ‘Who is Victor Hugo?’
Hallucinating AI: “The 19th century French writer Victor Hugo was also an enthusiastic innovator who contributed to the initial design of the International Space Station.”
Black box
Joseph Naklavo Nagula said, “This is a black box that is absorbing data. The result may be formally consistent and logically structured but they can actually have a true error.”
In addition to these real defects, AI ignores linguistic prosperity. Chattabot, Cubec French or Haiti and French regional pronunciation and Caribbean fought against the language of languages, such as Crooel.
The French presented by AI often show the purity of the language and the lack of stylistic micro.
Joseph Newclavo Nagawola Chutki says, “If Molier, Leopold Sadar Sidor, Ame Sisar, Mongo daughter learned how AI was writing in French, all these animals would start participating in their graves.”
This problem is especially more serious in multilingual countries. For example, diplomatic Joseph Nulbho Nagula was born in Cameroon, usually speaking in corpolis, which is a mixed form of French, English, Pizin and native language.
He said “I don’t think II is asked here in the Camplis, but they will get some meaningful answers” (I love this country) or “Retrunds-Moi Sharp-Sharp” (quick answer to me) will mislead AI models.
La Frankophoni’s promotion
The Joseph Unit for the use of Frankophony, La Frankophney, Frankophoni language represents 93 states and more than 32 million people worldwide for use of governments and Francoophoni languages. This company has focused this linguistic difference to its digital technique.
This group’s attempt was converted into the UN Global Digital Compact of the previous year, a structure adopted by the member states of the AI government. Since 2023, La Frankophoni, La Frankophoni, took advantage of its diplomatic network to ensure that linguistic diversity could become a fundamental principle in the creation of AI policies. The network also includes a group of the UN influential Frankophone ambassadors.
On the way, many unexpected colleagues also came. Lusophone and Hispanic support groups also joined the war. Even Washington supported him. Joseph Naklavo Nagula said, “America supported the inclusion of language in AI development.”
His efforts have got the result. Cultural and linguistic diversity is clearly recognized in the latest global digital compact. This is something that was initially buried in widespread discussions about accessibility. “Our goal was to bring it forward,” he said, “
This movement also reached the Silicon Valley. In September 2021, the Global Digital Compact, where the Global Digital Compact was officially adopted, emphasized the need to provide global knowledge access to many languages under Sundar Pichai AI.
He promised – “We’re working in the world’s one thousand most spoken language.” He again repeated this promise in Paris.
Global digital compact limit
Despite this facility, the challenges are intact. The biggest challenge of these is visibility. Joseph Wars Naklavo Nagula, “Frankophone material is often pressed by platform algorithm.”
Streaming giants like Netflix, YouTube and Spotify prefer popularity, which means search results in the English language elements.
He said, “If linguistic diversity is truly taken into consideration, a French -Speaker user should be considered to be the top French language film on its recommendations.”
There is one more barrier that has been ignored by compact and it is an unprecedented domination of English in AI training data. There is no mention of the conference on UNESCO cultural diversity in the compact, and Joseph Nakulbulo Nagula believes that this negligence should be improved.
Joseph Naklavo Nagula emphasized, “Linguistic diversity, La Frankophony is the spine of digital advocates.”
Giving the speed of development, bringing these changes should not be delayed.