AI: the ability of machines to learn human language
Posted: Thu Dec 26, 2024 4:40 am
The ability of software to understand human language and learn from a universal data source is no longer a utopia. And, like Bartolomeu Dias, it is now possible to turn the corner of fiction and discover new (programmed) lands, (intelligent) seas and (technological) cultures.
Artificial Intelligence (AI) at scale involves training large Machine Learning models to acquire the ability to solve global problems that can be applied in a wide range of domains , such as speech and image recognition . In contrast, more specialized models are trained to solve very specific problems with a more restricted use.
For companies that keep up with the growing technological and digital evolution, scale models can mean a revolution in the way intelligence is extracted from business data and how increasingly intelligent functionalities are built.
AI: the evolution of Artificial Intelligence
In the early 1990s, when Bill Gates founded Microsoft, he made a claim that was kazakhstan whatsapp number database credible at the time: “ computers will one day see, hear, talk and understand human beings ”. Visionary ideas are always debatable until they become a reality in people’s lives. This is what has happened with the evolution of AI over the last decade.
This evolution, marked by relevant events and others that may go unnoticed by those less attentive in the field of AI, represents an event as extraordinary as that of Bartolomeu Dias, in 1488, when he rounded Cape of Storms and became the first European explorer to open the way to other lands, seas and cultures.
The first landmark moment came in 2014 when Eugene Goostman passed the Turing test. Eugene Goostman was the first chatbot to pass the Turing test, which was designed to assess a computer's ability to communicate with a human in an "indistinguishable" manner. In June 2012, at the event marking Alan Turing's 100th birthday, Goostman won the competition billed as the largest Turing test contest of all time, successfully convincing 29% of judges that he was a human. In June 2014, at the contest marking the 60th anniversary of Alan Turing's death, 33% of judges thought Goostman was a human. This event validated Alan Turing's prediction in his 1950 paper "Computing Machinery and Intelligence" that by the year 2000, machines would be able to fool 30% of human judges after five minutes of conversation.
Along the way, we also saw the recognition of objects in images , the transformation of voice into text and the interpretation of commands, the translation of voice into several languages in near real time , Machine Learning services and platforms with widespread application and the enormous contribution of the Cloud and Mobile that have boosted all these events through the possibility of mass use of these technologies.
This path was taken by the major technology companies of the time, such as IBM, Microsoft, Google, Apple, Facebook, Amazon and many others. These technology companies created the services and platforms that companies use today to solve problems in the areas of knowledge in which they specialize.
However, the main event, the one that opens up more and greater perspectives regarding the future of AI, is the one that was recently presented by Microsoft at Build 2020 and that IBM has also been working on since 2011 with the famous Watson: the machine's ability to "understand" human language and "learn" from a universal data source.
It may seem trivial, but it is far from it. It means an opening up of lands, seas and cultures comparable to that of Bartolomeu Dias in 1488. It is simply a matter of giving the computer the task of "reading" a truly large quantity of "books" on a given field of knowledge and, after a few hours, expecting it to have learned the subject and be able to naturally answer the questions that human beings ask it using natural language. Is it possible to imagine something like this? It is like teaching a person from childhood to adulthood, from primary school to university, but in just a few hours. This is fiction that has already been seen in films and considered an extraordinary utopia!
AI at Scale in the Business World
How can companies take advantage of these trained models at scale? Could these models learn from a knowledge base of facts from the day-to-day life of a company? In the field of companies specializing in management software, what could ERP software solutions offer with this technology?
Taking PRIMAVERA as an example, and assuming that a scale model would exist today, when an invoice is issued through the system, the values associated with the transaction and the references for the entities involved are recorded in the database. But this transaction could also give rise to a sentence in natural language, created by an automatic process.
Basically, it would be something like: "In June 2020, customer Rosa Maria, from Coimbra, purchased a "Children's Wardrobe" item worth €179, two "Children's Bed (..)" items, worth a total of €550". For a Human Resources use case: "João Faria's remuneration was reviewed in March 2019 with a 5% increase". This process, applied to several areas, from Human Resources to Customer Care, would gradually drive the story of each company, customer and employee.
This information collection corresponds to the data that the scale model needs to learn "everything" about a company or a person, so that it can respond later in the context of the ERP functionalities. For example, when the user is trying to interpret a dashboard and wonders what happened in June 2020 to cause sales to be so low, the machine could respond with "In June 2020, the largest sale was made by the customer Rosa Maria, from Braga, for a total value of €550 and by 10 other customers for a total value of €1250. At that time, the world was facing the biggest pandemic of the century, due to the COVID-19 virus, which forced countries to suspend their economic activity."
The last part of the explanation can be achieved by feeding the knowledge base with external information from the political-social context. The possibilities are immense.
In conclusion,
Watson's performance in the Jeopardy finals in 2011 or, more recently, the fabulous explanation of AI at Scale, presented at the Microsoft Build 2020 event, are examples of the ability to apply AI in the business world.
Artificial Intelligence (AI) at scale involves training large Machine Learning models to acquire the ability to solve global problems that can be applied in a wide range of domains , such as speech and image recognition . In contrast, more specialized models are trained to solve very specific problems with a more restricted use.
For companies that keep up with the growing technological and digital evolution, scale models can mean a revolution in the way intelligence is extracted from business data and how increasingly intelligent functionalities are built.
AI: the evolution of Artificial Intelligence
In the early 1990s, when Bill Gates founded Microsoft, he made a claim that was kazakhstan whatsapp number database credible at the time: “ computers will one day see, hear, talk and understand human beings ”. Visionary ideas are always debatable until they become a reality in people’s lives. This is what has happened with the evolution of AI over the last decade.
This evolution, marked by relevant events and others that may go unnoticed by those less attentive in the field of AI, represents an event as extraordinary as that of Bartolomeu Dias, in 1488, when he rounded Cape of Storms and became the first European explorer to open the way to other lands, seas and cultures.
The first landmark moment came in 2014 when Eugene Goostman passed the Turing test. Eugene Goostman was the first chatbot to pass the Turing test, which was designed to assess a computer's ability to communicate with a human in an "indistinguishable" manner. In June 2012, at the event marking Alan Turing's 100th birthday, Goostman won the competition billed as the largest Turing test contest of all time, successfully convincing 29% of judges that he was a human. In June 2014, at the contest marking the 60th anniversary of Alan Turing's death, 33% of judges thought Goostman was a human. This event validated Alan Turing's prediction in his 1950 paper "Computing Machinery and Intelligence" that by the year 2000, machines would be able to fool 30% of human judges after five minutes of conversation.
Along the way, we also saw the recognition of objects in images , the transformation of voice into text and the interpretation of commands, the translation of voice into several languages in near real time , Machine Learning services and platforms with widespread application and the enormous contribution of the Cloud and Mobile that have boosted all these events through the possibility of mass use of these technologies.
This path was taken by the major technology companies of the time, such as IBM, Microsoft, Google, Apple, Facebook, Amazon and many others. These technology companies created the services and platforms that companies use today to solve problems in the areas of knowledge in which they specialize.
However, the main event, the one that opens up more and greater perspectives regarding the future of AI, is the one that was recently presented by Microsoft at Build 2020 and that IBM has also been working on since 2011 with the famous Watson: the machine's ability to "understand" human language and "learn" from a universal data source.
It may seem trivial, but it is far from it. It means an opening up of lands, seas and cultures comparable to that of Bartolomeu Dias in 1488. It is simply a matter of giving the computer the task of "reading" a truly large quantity of "books" on a given field of knowledge and, after a few hours, expecting it to have learned the subject and be able to naturally answer the questions that human beings ask it using natural language. Is it possible to imagine something like this? It is like teaching a person from childhood to adulthood, from primary school to university, but in just a few hours. This is fiction that has already been seen in films and considered an extraordinary utopia!
AI at Scale in the Business World
How can companies take advantage of these trained models at scale? Could these models learn from a knowledge base of facts from the day-to-day life of a company? In the field of companies specializing in management software, what could ERP software solutions offer with this technology?
Taking PRIMAVERA as an example, and assuming that a scale model would exist today, when an invoice is issued through the system, the values associated with the transaction and the references for the entities involved are recorded in the database. But this transaction could also give rise to a sentence in natural language, created by an automatic process.
Basically, it would be something like: "In June 2020, customer Rosa Maria, from Coimbra, purchased a "Children's Wardrobe" item worth €179, two "Children's Bed (..)" items, worth a total of €550". For a Human Resources use case: "João Faria's remuneration was reviewed in March 2019 with a 5% increase". This process, applied to several areas, from Human Resources to Customer Care, would gradually drive the story of each company, customer and employee.
This information collection corresponds to the data that the scale model needs to learn "everything" about a company or a person, so that it can respond later in the context of the ERP functionalities. For example, when the user is trying to interpret a dashboard and wonders what happened in June 2020 to cause sales to be so low, the machine could respond with "In June 2020, the largest sale was made by the customer Rosa Maria, from Braga, for a total value of €550 and by 10 other customers for a total value of €1250. At that time, the world was facing the biggest pandemic of the century, due to the COVID-19 virus, which forced countries to suspend their economic activity."
The last part of the explanation can be achieved by feeding the knowledge base with external information from the political-social context. The possibilities are immense.
In conclusion,
Watson's performance in the Jeopardy finals in 2011 or, more recently, the fabulous explanation of AI at Scale, presented at the Microsoft Build 2020 event, are examples of the ability to apply AI in the business world.